The mood at this VON was not like other Pulver events I have attended. Even the enthusiasts were a bit more subdued as the industry downturn clearly hung over people. I'd say the crowd was divided into 3 categories:
· For those who sell into the US carrier market, the mood was generally grim. Lots of gallows humor in the talks about "Telecom Nuclear Winter", or "Telecom Survivor"
· For those who sell to international carriers the mood was a bit more upbeat. People are buying. IP Telephony is becoming normal, gaining share, and generally indistinguishable from the PSTN. Carriers are buying it because it's good enough, cheap, and getting better – a classic "disruptive technology". Nobody wants to buy the last circuit switch off the production line.
· The brightest spot was for those selling to enterprises and selling services directly. The deployment of Windows XP and the growing influence of SIP create great reasons for optimism. Big markets driven by new services – definitely not PSTN telephony.
The industry view on CALEA, 911, and other safety/security services has changed dramatically from what I recall. Instead of being seen as unrealistic government intrusion, this is seen as necessary security. The suvivability of IP networks over traditional PSTN is now seen as a big selling point.
Microsoft is of course very big. One thing I found interesting was that the rather than looking revolutionary, the MSN version of IP telephony looks like an evolutionary normal service. Like buying
minutes for a wireless prepaid service.
Presence changes everything. Microsoft has bought into presence as the unifying force in communication and SIP as the enabling technology for IP Communications, not just voice telephony.
IP Phones are multiplying. They are still too expensive, but that now means $300, instead of $500 or $700. I see no reason why a smart, full feature IP phone won't become more than competitive
within a year.
Finally, I came to a personal realization here. Maybe it took being out of a major equipment vendor for 3 months to let myself think like this, but during the 3rd day of the conference I had a sudden realization that IP communications is not likely to be an incremental technology introduced into the PSTN model of
communications the way stored program control, Digital switching, and dozens of others have been, but is instead likely to be a completely different animal. More like the introduction of the telephone in a world where people used to depend on the telegraph and paper mail. I believe it is a technology that won't directly replace the PSTN, but instead will create an alternative multi-media communication service that may well take traffic from the PSTN.
The PSTN won't go away, but it will be difficult to introduce the same technologies and services within the constraints of legacy technology, operations, and regulations under which most PSTN
carriers currently operate. The scenario for introduction of a new communication services network is basically:
Rapid deployment of inexpensive or free PC based clients, augmented by increasingly affordable IP phones and wireless clients creating a large base of potential users.
Continuing growth of broadband connectivity to both office based clients and home based clients
A significantly better communication experience coupled with "Metcalfe's law" and peer pressure driving more people to connect.
All of this creating a robust market in services, but primarily in the end-to-end IP space (including endpoints
like wireless phones with IP data and circuit voice).
The PSTN does not go away in this scenario, since it does things the IP network won't necessarily do. It also reaches the people who are not connected, and just as I'm sure there was a business in delivering phone messages to people who did not have a phone, there will be a robust business in interconnecting the PSTN and the IP communication worlds.
Here are my more detailed notes on the experience and the sessions:
This was my first conference since leaving Lucent, and my first trip away from home since 9/11/2001, leaving me more than a bit uneasy about exactly what to expect and what I'd find had become of the industry. In the days before VON I had been contacted by a couple of people who got my name off the VON email list, looking for jobs or partnership. There are clearly lots of people in the industry who have been displaced, and VON was a focus for them to get reconnected. My airline travel was uneventful, except for the long delay going through security. Judging by the number of people in the airport and on the plane I'd say our fear of flying is rapidly fading.
Arriving at the show it took no time at all to see some familiar faces. Everyone speculating on what and who may or may not show up at this conference. One told me that 70% of his contacts are now self employed. The only thing we all agreed on though was the industry would improve, just a matter of time.
Most of the familiar faces showed up, but a lot of familiar companies had reducd or even no presence. The attendance was probably less than Spring VON, but still a good crowd.
The Venue for the show (the Georgia International Convention Center) was near the airport in an area with many hotels in walking distance, quite convenient, except for the fact that the area has few restaurants. Facilities were as usual first rate, with good network connectivity, wireless LAN, and excellent food and refreshments, with plenty of time in breaks and receptions to network with other attendees.
I attended this session for the morning and most of the afternoon. This one had a pretty good crowd, probably 100-200 people, almost a full room. Attendees included Henry Sienrich, Scott Petrach, and a few other notables. When Microsoft talks, everyone listens.
2-3 years ago we were in the PSTN Communication era:
Innovation endeavors:
"The PSTN is at the end of its innovation cycle -- There's nothing more to do with it". (Comment -- I'm not sure I'd with this in there is a lot of room for innovation with things like PINT, SPIRITS, and Parlay, but carriers are not moving rapidly to introduce these kinds of interfaces and without them innovation is slow.)
Communications today: Silos with dedicated services and infrastructures for different servies.
Microsoft's vision: unified experience via many devices, with different media but common standards based infrastructure. (Comment -- in other words a lot like pictures shown in many previous attempts to unify communication.)
Presence based multi-modal communications is the new era of communication, a revolution (Hmm, where have I seen this before. The vision is almost exactly what I was showing at The Presence/IM conference a year ago). This includes telephony (circuit and VoIP) collaboration, and other applications.
The core is presence enabled session management. Session is what makes the differnece between single mode and multi-mode communication. The key to be able to integrate the multiple infrastructures.
Why SIP? Microsoft was working on it 21 months ago. What they saw in SIP was simplicity for rapid innovation (like HTTP), multi-modal session support, and readily integrated into the industry. SIP isn't complete. He talked a little about the history, citing PINT as one of the reasons they got interested in SIP. He talked about the success of
SIMPLE (SIP for presence) including broad adoption.
The .net vision: End to end networking (a network operating system) integrating software on different devices. How does .net integrate with real-time? Multiple clients, not just PC. They interact through services with servers. .net provides the overall framework.
Question: "What's the difference between this and the telco vision of IN 15 years back? Nothing. This looks like IN (Comment - Actually more like TINA. 15 years back of course neither was around, but the vision was).
What's the difference? Today we have the devices to make
this happen which wasn't the case before.
(Comment -- yes, this does make a big difference. I worked on intelligent devices and network
environments in 1978, but the cost and terminals weren't up to delivering it.)
He talked about setting up an ad hoc network by using 802.11 equipped PC's in point to point mode with SIP servers setting up an ad hoc network for communications. Couldn't do it 2 years ago but now we have the software for it.
Up until now interpersonal communication with the PC is text base (Chat, IM). We have not achieved the potential of the PC to do more (Voice, Video, Application sharing, etc.). We needed a new platform (Windows XP). XP is based on NT, no DOS, more reliable than any other Windows OS. "Windows XP is just as good in reliability as any Unix machine I ran 5-10 years ago". (Comment -- this is an interesting phrasing. Unix has of course gotten a lot more reliable in that time.) New user interface that is task oriented.
Windows Messenger: This is the unified user interface.
They made a big pitch for interoperability. The have lots of SIP equipment in their lab and have been testing interoperability. Claim this is much easier integration than they were doing with H.323 (No ASN.1 among other things).
What's the server story? .Net RTC Server. It will work on Windows 2K and Windows .Net server (new) -- a SIP Proxy and registrar, directory and secure federated enterprise IM solution. (Comment -- A natural fit for Microsoft's strengths). They have a partnership with Reuters and 21 investment banks deploying this. The have lots of pull from wall street firms for SIP solutions.
.Net services -- Passport and Messenger are part of the .Net lounge 40M users on messenger, and 120M unique users on Passport (Comment -- of course you can't avoid Passport if you do anything with Microsoft's web services.)
He talked about Halistorm and the storing of personal data on Microsoft server centers. Today's passport and messenger notifications are pre-hailstorm. Version 1 will ship in 2002. It will have APIs and XML based presence queries.
XP will go onto 120M desktop PCs in the next 18 months. David says his performance review depends on making their platform work with ISVs. Talked a lot about gaining the trust of the ISV community.
The real-time Client: XP and CE 4.0 will be first. The server (Whistler) -- SIP registrar and Proxy, Extension APIS. The .Net RTC Server (end of 2002 Beta in March): A solution for corproate IM and Presence.
Question: At what milestone will we be able to take an XP client and interwork with a Dynmicsoft or other SIP server for registration, etc. Answer: December 15th. Henry Sinnreich -- Actually the SIP voice client works with other proxies today. David -- Yes, he was talking about presence and IM, for which interworking is scheduled only in December.
Some more thoughts on XP: Microsoft and others will spend $1B promoting it in the next 2-3 months. He believes that many more home PCs will have it in the near term than corporate PCs.
The strategy: Use the rollout of windows XP and the windows "footprint" to drive the deployment of a common platform for real time communications.
We are getting to the end of traditional productivity tools (e.g. Office). "How many more features can you put into word?" In response they are focusing on expanding ideas for productivity. One thing they are working on is a tablet PC -- he's a skeptic, but has been won over by this. It replaces the paper pad by recognizing handwriting on a pad and making everything digital.
Communication productivity on the other hand is just beginning -- what we have done up until now is very basic -- dial a phone and hope it rings. (Comment -- yes, this sounds like a great story, one I have told, but it has been very hard to convince a carrier to spend money on equipment to realize it) Some things we want to do with communication productivity:
Some ideas from the past:
The goals going forward:
What are they building:
A note on the meaning of Offline presence: This is the persistence of requests to watch presence. It doesn't look like it actually holds your presence.
Presence enabled client:
Question -- what's the transition away from TAPI? Not clear. The real issue with TAPI is they aren't investing in updating protocols. Not clear whether they will put a TAPI-like API on top of SIP.
Lots more questions about the planned phase out of TAPI, Exchange based IM, etc. Their philosophy is clearly to push users as fast as possible into XP and Messenger, but it's not clear how well that is playing with the enterprise community that has a big investment in the older solutions.
Question: How does interworking of IM clients work: MSN clients will go through the MSN cloud (unencrypted) as they do now. SIP clients will go peer to peer using SIP, or via Proxies if record-routing is used.
Clients can have multiple registrations with different servers.
Some demo capabilities on ease of use:
NATs -- They have done work for home PCs sitting behind NATs to have the PC request a port assignment. Given the port number and real IP, the PC can send the right information in an invite to get the media set up correctly. There is a good deal of support in the industry for this kind of request (UPnP -- Universal Plug and Play). Microsoft XP can be used as a NAT (use 2 network interfaces and software). They have signed up all the industry leaders to implement this. It may need a firmware upgrade for some of them. They hope that broad deployment will put pressure on non-compliant NATs to comply.
Instant Messaging - People discount the importance of IM in touting Presence as the realy need It is a persistent killer application independent of presence and independent of voice/video, etc. Big advantages: Quiet (use anywhere). Lightweight (doesn't need broadband) (Comment -- it's also parallel. You can chat with multiple people via IM, don't try this with audio!)
Audio/Video -- Echo cancellation, Delay -- it's really 120ms, not 70ms when you add in network delay and echo cancellation. (Comment -- he said this is well within the limit of good quality speech. Maybe, but my experience is 120ms is still noticeable. That's a quarter of a second round trip!) Default codec is a 16Kb/s G.722.1! (comment -- yes, this would make a big difference in perception, and maybe even compensate for the delay). They optimized everything for cubicles and small office space.
"SIP is it" -- It unifies computer and PSTN. (He really said something about breaking the carrier network model).
IM is here to stay, telephony is here to stay, SIP is the bridge that unifies them. Vendor CTI? -- the slow road to nowhere! Been around forever but going nowhere. SIP is where it will happen.
Question on 3rd party call control (3PCC): Mid December version has PINT in it, but not SIP based 3PCC.
Traditional telephony -- Telephony is really an island, not connected to the data world. May have an asynchronous bridge in the form of a unified messaging service.
A Integration scenario on how a business would use this with the PSTN: PBX connects to a gateway (generic), which sends to the RTC server (SIP based). All the smarts are in the RTC server, which provides the features. He showed a scenario where a call may get bounced back into the PSTN to be routed to a mobile phone. (Comment -- sure, but you have now hairpinned all the way into the RTC server, through the gateway, PBX, and trunking, not likely to be cheap).
Scenario -- recieve a call when in a meeting -- use IM to send a message to the caller. (Comment, another one I was showing a while back. Difference is I was pitching this as something you wanted to do inside the PSTN via VoiceXML, Microsoft doesn't want to touch the PSTN and does it on the internet side. Fine, except again you burn up expensive PSTN interfaces to do it.)
Federation -- ability to form communities of trust and share presence and contact information. They want to enable this ad hoc or via central servers, use a directory to locate a potential buddy, etc.
XP+++ (Futures)
Windows community: 110 M using Windows 95, 240 M users using WIndows 98, goal is 100M Windows XP users within 18 months. Given that they are porting the real time stuff to 98, that's 340M potential users (Comment -- I wonder how many of these are unique people.)
The opportunity for the VON community is to build applications on the APIs in the client and server.
Service Provider opportunities for Microsoft RTC:
Microsoft feels their key assets are: MSN, their brand, hailstorm (now "My Services") The XP SIP Communication Client and the largest developer community.
The Telco Service Provider assets are: Subscriber base, the network, voicemail and enhanced services, Billing!, brand, location and presence information, management network with guaranteed QoS.
"Every shipped copy of XP with RTC Client is a handset in need of a carrier". (Comment -- yes, but if they get used primarily for PC to PC you don't need a carrier!)
On 10/11, they announced some new relationships with carriers, including Telus and several IP carriers. Coming is BT and another major international carrier.
Opportunity is for carriers to terminate "PC to Phone" calls initiated from the MSN client. (The slides talked about things like TCP based protocols that would traverse firewalls using HTTP. I presume that's not for the audio!)
The user signs up with a carrier, this is maintained in the .net servers and authenticated using passport. Calls are forwarded to the carrier. Microsoft will generate call detail records for all the calls. (Comment -- It's not clear from what was presented whether this applies to ALL calls or only calls that terminate outside of the internet -- i.e. do you need a carrier (and pay for minutes) if you terminate onto another XP client? Obviously not, but that may be the way it's architected)
When you make a call and haven't picked an ITSP, Microsoft offers you a page of their partners in random order.
Question -- does the service provider have to be one signed up with Microsoft? Answer -- didn't answer, said that they were asking carriers to meet certain standards then they get to be part of the MSN listing. Client can talk to anyone, but it sounds like provisioning it is a pain. Lots of discussion on this. The answer seems to be that the client will support a number of different services, including raw SIP. The Passport based MSN calling service is the one that requires you to use a registered carrier. (Comment -- This looks like a "normal" phone service, not IP telephony. It looks very evolutionary, like signing up for prepaid and refreshing your account).
Question (diatribe?) Why tie this to an operating system? You won't push XP into every device. Answer -- the Infrastructure is standards based, there are clients for Macintosh. Microsoft does sell operating systems though if you want them.
Question: What's the ITSP interface based on? Answer -- It's a Microsoft standard (Comment – This didn’t provoke much of a response. I presume the only thing Microsoft did was define some XML based interfaces for obtaining billing information and the like and they have no long term interest in keeping it proprietary, there just aren’t good standards yet in IP for carrier interworking )
Question: What's in
it for the carriers? Looks like just a
pass through that could get in the way.
Answer -- extensible client with good user interface on everyone's PC. Service providers can use the Client APIs to
extend the client User interface. (Not
initially, beta now). Microsoft provides a standard IP telephony experience. (Comment
-- this is very much a good news/bad news situation:
Carriers can feel free to develop their own version of any of this. IM clients, servers, etc. Microsoft feels though that clients and web based servers are what they bring to the relationship and the value they add.
Question: Take the example of Delta 3 -- have their own client and servers, can they take the APIs and build their own branded service completely separate of MSN -- Answer -- yes. Henry Sinnriech: two people on a LAN in Singpore should be able to communicate without benefit of carriers and proxies, but that's different from this.
Question: Is the Client kit free? -- yes, in spite of the fact that the beta program had an awkward registration process. That will go away.
Business terms: Microsoft brands the client with the carriers brand when you sign up. Microsoft will market the services. The carrier owns everything else. Microsoft asks a one time fee of $100K for a carrier to sign up to get access to their SIP infrastructure. Messenger client and service will lanuch 10/25/2001. http://www.microsoft.com/serviceproviders/voiceservices for more information.
Question: How will the phone to PC part work. Answer -- too preliminary to tell.
Question: How many SPs are signed up. 4 so far, 2 more coming up by the launch of XP. 2 more in Japan coming up for the launch of XP in Japan. Working with lots more.
Talked about the signaling interfaces and models including a PINT API for connecting phone to phone.
Acoustic echo cancellation -- really works only on your end of the call and built into the OS, models the sound return from the speakers. Works well only in small rooms. (Comment -- echo control gets harder and harder the longer delay there may be in echo return. This sounds like it would work okay for PC to PC calls but won't do much for PC to phone calls where the echo problem is coming from the far end).
Auto Gain Control -- Nice feature, but it doesn't work when echo control is on and vice versa. (Comment -- yes, that's why good speaker phones and conferencing microphones aren't cheap, it's a hard problem to do both.)
Auto adjusting of codecs and channel capacity based on detected bandwidth. This is an advance over netmeeting where this was set manually.
Long list of selected codecs. IM is via the draft SIMPLE standards. Windows XP supports only point to point IM, the SKU server will support conferencing (future).
Question: Can someone plug in a 3rd party codec (not now, but considered for the future, though their plan isn't necessarily to do this. They want to provide all the ones that you really need themselves and avoid the need to support a pluggable codec API. They can do this in Windows CE).
Provisioning: Out of band (XML pushed through the network to the client), Server based provisioning -- when client registers the server returns provisioning information.
Policies: Allow administrator to control the media that users can use (i.e. limit video, restrict IM communities, etc.).
Profile: Has service specific information for use by the service.
(I missed some of the API session here while attending some of the VON Demo track, and rejoined this later for presentations by their partners who were working with them using the APIs.
Ubiquity -- They built an end to end SIP solution using an application built on the XP client. Worked only with the PC to PC function from the API, configuring it to use the Ubiquity proxy server. 4-5 weeks worth of development using C++ (Visual Studio) COM based APIs.
It has a simple user interface with windows for contacts, video, and actions (audio tuning, click to dial, send an IM, etc.) Client could add buddies and display their status, and each buddy has menu for calling and IM. He set up two phones so they were on each other's contact lists. Then could make a call between the two (Voice, video, and shared whiteboard). Application sharing (shared a powerpoint application), like Netmeeting. Interestingly enough the time lag between changes on one PC to the other was very apparent, even though the PC's were plugged into the same LAN hub. It may be that the video was flooding the local network.
CosmoCom -- An IP call center that routes IP telephony, chat, email, etc. to call center employee's multimedia PCs. They are an older company in this business (1996) and are targetting major service providers.
Started on Netmeeting, moved on to build on TAPI, and now on RTC. They are hearing demand for SIP from telcos and customers. The voice comes in through a VoIP gateway, and then they provide the IVR function in VoIP in their server to route the call to an agent. They recommend a managed IP network dedicated for this. For IP contacts, they can query whether the user as XP and if so can set up voice and video connections.
Quick prototype for the client took less than a day. (Very simple stuff, just the voice connectivity, they plan to add presence and central provisioning).
One interesting comment was that they observed a great reduction of call setup up time versus using H.323 (hundreds of ms versus 2-4 seconds).
This was very thinly attended. Maybe no surprise, the attendance of this track was light in Phoenix as well, and it was a great weather day.
Congruency -- showing a unified communication application (IP Centrex, with voice mail and other capabilities.) They license it to enterprises. They used to be an ASP but abandoned that model.
What was demoed was provisioning of new users, voice mail, and other basic services with both web interfaces and a SIP phone interface.
DynamicSoft -- Demoed a "wake up call" service.
The Internet is more than just transport, it's a collection of applications. Real objective is to create new applications that put these capabilities together to create new services. The Application Engine is Dynamicsoft's product for this. Target market is for Solution vendors and Carriers.
Servlet based APIs. ("Any recent CS graduate will understand them"). Abstract from SIP -- Developers don't need to be SIP experts. (Comment -- interesting. 4 months ago Jonathan was pitching how essential it was for the applications to be aware of SIP to use its capabilities. I guess you can be SIP aware without being into the details).
Network Decomposition -- Instead of one big server with everything together, components connected via protocols. The application is central, but doesn't do everything. Centralizing the application in one place makes upgrade and programming easier, but distributing resources allows multiple applications to leverage them. More scaleable than a monolith.
Demo -- used Outlook calendar to enter a schedule for events. Then used a web page to enter where the wakeup calls would be made to, and information to be delivered. This gets delivered to the application engine which delivers his message, the weather (pulled via SOAP from a weather server), and data from the calendar (using Exchange code that they built).
The Demo used a Nuance based VoiceXML server to deliver the audio, and the Application Engine is using 3rd party call control to initiate a call between the media server and a PSTN gateway used to call the actual phone.
Application that they showed was developed in 1 week by an intern. (Comment -- Not bad, but I have to wonder whether it handled any error leg cases).
Other 3rd party call services: Auto conference, traffic jam (call me if there is a traffic problem), many others.
Pingtel: They had 3 phones, two hard one soft, plus a PSTN connection via a gateway somewhere (not clear where.) They demoed a LOT of services:
Question: Do the services live in the phone or in the network? Answer -- both. They are targeting enterprise customers where the services may be in enterprise servers.
Question: What's the business model, who sells this stuff to who? Answer. The solution provider will provide the pingtel phones and applications. Often this is done by the service provider (ITSP or PSTN).
IpDialog: They provide software building blocks for OEM solutions. They were demoing "SIPTone, a cheap SIP phone. Full features, low cost. Low cost is critical to ubiquity. (Comment – Yes, eventually, but the real need is to get something attractive enough to get everyone to buy into it. Economics of scale will make just about anything cheap enough if millions get sold.) They also sell H.323, MGCP, and Megaco phones. They are selling to ITSPs (provide a phone with an account) and IP PBX makers.
Configuration: Manual or DHCP for network stuff, phone identification, registration, and proxy (Comment -- This was a pretty good sales pitch for Pingtel and others that want to push all this stuff from the server. This isn't something an end user wants to do).
Showed a simple call (no audio). Showed an IP message trace of the SIP and RTP messages. This was clearly a demo aimed at engineers.
Their market research shows that cost is the most significant barrier to deployment. Everyone thinks VoIP is cheaper, but not when the phones cost $300. They idea they have is that at least half of the phones in an enterprise don't need features (conference rooms, people with limited feature needs, etc.), so by mixing in "dumb" phones with the smart ones lowers the average cost per station. Also made an argument against putting all the features in the phone, since when the feature is in the phone you pay for it whether or not you use it but in the server you can scale capacity to support only as much as you need.
Global IP Sound. "Better than PSTN over the public internet"
Demo setup using the public internet to Norway, to someone sitting on a DSL line at home. This is the 16Khz codec that Microsoft was using I believe. The voice quality was very good. The delay wasn't too noticeable in the demo, but there wasn't a lot of tight interaction between the two. There was no apparent echo. The audio was apparently two way, though (we generally heard nothing while the the person on this side was talking). The big innovation seems to be an improved jitter buffer that adapts quickly to increase and reduction in delay.
They ran a route trace on the connection and had at least 10 routing hops and 350ms in round trip delay in the connection.
A good crowd of people, but clearly smaller than Spring VON. A different one too. More small companies, and a higher proportion of people wearing suits than jeans. I'm not sure what that means for the industry.
Jeff Pulver -- Jeff struck me as a bit lower key than he has been in the past. He thanked us all for coming in spite of the climate. People from 32 countries came, over 575 companies (55% vendors, 25% service providers, 20% other (consultants, VCs, etc.)) Most raised their hands when asked if they had been to a VON event before.
9/11 Terrorist attacks may accelerate the adoption of the technology:
Is the industry at a crossroads? Not according to Jeff, just at the beginning of the IP revolution. The last 6 months have been like VON Survivor, every meeting people look to see who got voted out. Incumbents have some immunity because they have customers.
He talked about the trap of replicating the PSTN in IP. "IP Dialtone" sounds good, makes people comfortable, but it's probably not right, dialtone is just wasted time, look for innovative new ways to use the technology. Don't just replicated 1960's engineering and human interfaces.
Business opportunities today:
9/11 events make clear the need for features like CALEA and emergency (911) services. (Comment -- very interesting, since 6 months ago many in the VON community were trying to avoid having to do features like this.)
It's too late to sell to incumbents. They like to have small number of long term vendors and they already have them. Seek out new opportunities. End-to-end IP is key. Remember that 60% of all communication is "local", that's an easier market to enter than being ubiquitous over the world.
He showed some interesting market data. Revenues at Cisco dropped from nearly $7B/quater a year ago to just over $4B, but the voice part of that had steady growth (hard to interpret his charts as to what that really meant, growing in absolute dollars or growing as a proportion of their total).
For Fiscal year ending end of July, Cisco sold >3.5M DS0 VoIP capacity (gateways). He had a market segmentation chart, roughly half and half service provider and enterprise, in each, 1/3 was access (terminals and local access, probably DSL), and 2/3 transport (PBX-PBX, IP Trunking).
IP PBX customers -- total is growing, but the number of new ones dropped back so the rate of new customers is about what it was a year and a half ago.
Service Providers -- They have 60 customers, the growth rate doesn't look like it has dropped (linear, not exponential).
Cisco has done a lot of work on International versions of SS7 (showed a big chart of variants, many of which I recognized from past work. Nothing obvious missing.)
The drivers for VoIP -- not changing very rapidly. He had a chart showing that in the long term the drivers shift from transmission savings and arbitrage to services and operations savings. The chart actually showed the relative importance of each of these factors over time on one graph . The interesting thing is these are all "hockey stick" curves, and we are in an era where all the curves are low. This suggests to me that this may be why things are slow -- we have used up the momentum in arbitrage and cost, but haven't yet reached the point where services and operations savings are compelling. Arbitrage is apparently still a big driver for some international markets.
Gateway protocol forecasts -- In transition from a 100% H.323 world to a mix. He is much less positive on SIP than he was before or even than others are now, sees H.323 and SIP coexisting for a long time. Asia is big area for deployment of gateways and they are H.323. The market may become bifurcated globally like wireless, with different regional preferred technologies. Also sees a growth in "MGCP" based gateways. (Comment -- I presume he meant Megaco, but I don't know since Cisco is certainly a leader in MGCP).
"Powered Line cards" an important trend in enterprises that is being mixed (I'm not sure exactly what he is talking about here, probably an interface to power a POTS phone.)
Retail and wholesale ITSP partnerships (showed Microsoft's Voice.net)
"SALT" -- Speech Application Language Tags -- This is their view of what VoiceXML should be. VoiceXML is very nice, but not well suited to multi-modal communication. SALT has a consortium and is working to build on VoiceXML, HTTP, etc.
Softswitch status: Tandem and PBX access market is maturing, lots of deployment, clear it works. Local is still early. CALEA and 911 service are key issues (Again, 6 months ago I think he would have said it was wrong to apply CALEA to IP).
(Comment -- It's interesting that his slides came up in the wrong size. He got them fixed, but it seemed to take a bit of time to do so. Not a good advertisement for XP).
6 months ago he listed all the usual drivers for Voice over IP, and said we were waiting for the revolution. Up to now it has been slow to materialize. He said the barriers were:
The applications are coming rapidly with SIP, for the infrastructure, we need to have a better user experience and the Client is key to it. The key is an integrated communication experience (not just telephony, IM, etc.), and Windows XP is there to provide it.
He talked about issues in the enterprise:
Home issues:
He talked a lot about NAT (Network Address Translator) issues. Characterized a NAT as the internet version of the old extended memory schemes for DOS (a reasonable analogy). Their solution is UPnP (Universal Plug and Play) This is a protocol scheme to open ports in a NAT dynamically. They have a standard, a consortium and commitment from 85% of the NAT vendors to it. Still they believe IPv6 is the real answer and may be forced in Europe and Asia before the US because of acute shortages of IPv4 addresses. Sees the adoption of the technology driven from the end.
Firewalls -- break UDP and incoming TCP. Not consistent with the internet model of transparency. They are working on "Firewall Control Protocol".
Sees big opportunities in infrastructure for media in networks. Lots of new things possible and not well defined.
This is the year of SIP. Big upswing in % of people now using SIP, but VoIP is less than 0.5% of world traffic -- that's a real opportunity, since 50% is a realistic guess for what's possible.
A big obstacle to introducing anything is multiple protocols -- you can fix that, just do SIP
He showed a p9icture with lots of SIP Phones (Not just Pingtel and Cisco any more).
Showed a Windows XP messenger screen and connection to PSTN via Commworks gateway -- (which he said Worldcom is using in their network)
SIP devices for mobility. Swisscom is implementing presence on existing phones.
Problems with SIP devices:
Consolidation of services on the web and internet He showed a slide with side by side protocol stacks for internet and PSTN. lots fewer protocols and more rational prototcol suite than telecom, which has lots of disparate protocols for different solutions.
Proprietary PBX versus SIP Compliant IP PBX (SIP is less mature, but does lots of things that the proprietary ones can't, interchangeable phones, mobility, productivity tools, adding applications, carrier service integration, etc.) "Merger proof" (Comment -- interesting thought. Mergers create real problems for telecom managers)
Also a long comparison with Softswitch and SIP services, naturally favoring SIP.
Two approaches to VoIP:
He then spent a lot of time on QoS:
Intersting perspective, he came from Cascade (Later known as Ascend and the Lucent).
Packet voice is rolling in slowly. Enhanced services roll out from the core of the network to the edge. Build the trunking first, then business, residential, and future applications later. (Interesting, completely opposite Jeff Pulver and Microsoft).
Major carriers deploying (included BellSouth and QWest, Level 3, Global Crossing, Williams and a bunch of others).
Some discussions on different carrier business models (peering, retail/wholesale, etc.)
This is all "Network Convergence" based on facility re-use. True convergence is at the service layer, and the softswitch will drive this -- all services built in a common distributed softswitch platform. (Comment – a natural view from a strong softswitch vendor, but others would agree on the importance of service convergence but put the services somewhere else depending on their business model)
The future of VON:
VON is alive and well. Carriers still moving to packet, solutions are here today. Value proposition is still strong. Voice over Packet will outgrow other technologies like Optics which are overbuilt.
He was the CTO of BBN, now a consultant to the FCC on VoIP. Michael Powel apparently asked to have an advisor on Internet matters, not because the FCC wanted to regulate it but because they wanted to know about it.
Regulatory challenges: Technology moves fast, Regulation moves slow. The act of 1996 takes a definitional approach to differentiate IP telephony and Telephony based on technology. This isn't particularly good because the definitions mean less and less. Regulation should be on services, but it can't always be.
IP telephony flourished without regulation and the FCC is happy about it. The main pressure for change comes from other countries, because IP bypasses state revenue sources from telecom monoplies. The US isn't all that sympathetic.
The Stevens report of 1998 presents a more recent view on IP telephony regulation:
IP Benefits the US by driving down international long distance costs for us.
The European commission issued a recent view on regulation:
This was reaffirmed in a statement in January 2001.
IP Telephony and the ITU:
ENUM -- This opens a pandora's box. Doing Enum is a collision between cultures -- internet and regulated, etc.
The US recommendation:
Broadband -- FCC doesn't know what to do with it. Voice over cable and cable access to IP and Voice over DSL look different to them but are clearly the same service.
Question (Teressa Hastings) -- Is an IP Phone a Phone or a Computer? -- Couldn't give a clear answer.
Question (Chet McQuade) doesn't segmenting the market into different regulation domains hurt value (Metcalfe's law)
Question (Carl Ford) When will we get ITU standard implementations of ENUM? ITU is moving quickly for ITU, maybe a year to 18 months for a standard, implementations are different, There are implementations now but a standard one needs government action and that's going to take 2 years or more.
He gave an energetic talk on services architecture.
You could put all the servers together to preserve a familiar model, but this really is a waste of resources because they don't scale the same way, and furthermore it creates thinking problems in that people still think like there is a central office.
The server components include:
Voice quality -- can be beyond PSTN. Conferencing may push demand for VoIP -- someone dialing in through PSTN may be embarassed by sounding bad. (Comment -- This would be interesting to see).
Build lots of nice new features but make them usable by humans. He had a great Dagwood cartoon about a new phone system that was too complicated or intimidating to use. Another example, why not use that 2 line LCD on the ISDN phones for help, not useless status information.
Question (Chet MCQuaide) -- with all the distributed implementations, how do you keep it simple to use? Joe's answer was good GUIs like windows, and use the PC that is near the phone. (Comment -- Another answer is that you really have no idea how much complexity is behind your favorite internet portal. Telephony needs the equivalent of "My Yahoo" )
Session: Location Services –
This was a very lightly attended session, not more than about 15 people.
Today we have cellsite/sector location. E911 mandates lattitude and longitude, and asked for altitude (Z). Z was not accepted but new push for it after the WTC attack.
Technologies for location:
He showed a slide with market projections over time from various pundits. This shows a market of $5B-$20B in the next 5 years depending on who you believe. But lots of work left to do to get any of that market.
Where is the big market:
Small company productivity services:
Basic message was you need high accuracy technologies (GPS, EOTD, etc.) because you have to exceed the initial expectations of the users, which are high because of hype. Cell and Sector won't cut it.
Location is really about context. Context inicludes awareness of where a user is, what they are doing, what they are trying to accomplish, and the environment (terminal, network). All this should alter the behavior of the application. Mobile computing is a challenging environment -- bad networking, bad terminals, poor usability. Context is the way you work around this.
Session on SIP.
This was better attended, maybe 50 people. It focussed on the SIP standards working groups, with each speaker covering a different group.
Jonathan Rosenberg (Dynamicsoft) -- Covered the SIP Working Group. This one covers the core of the protocol. He went through a whole list of different things in the works which include some security, reliability, subscribe/notify, a Conditional connection capability to require the destination to verify the media path before an invite becomes valid, and several more. Jonathan’s material was a straightforward summary of the issues the group has been working on.
SIPPING is a gatekeeper for extensions. Basic purpose is to filter things out and determine whether the requested extension is really a good application of SIP, and if so whether what is needed is an extension or simply documenting a way to use SIP.
SIPPING has a lot of work on PSTN use of SIP, including putting in requirements for 3G. Also documenting basic call features, Forward and Transfer. Some submissions on conferencing, but nobody is really pushing this forward hard. SIPPING has a 200 page draft explaining how to do various PSTN services in SIP. ALso document on how to do T.38 FAX services using SIP.
Now there is a lot of support for the view that SIP will become the protocol for instant messaging and presence. SIP already provides subscribe/notify needed for monitoring, the MESSAGE method is a needed extension, but not difficult. SIP Registrars become presence servers.
"IM is a gateway drug for telephony" IM is a pure interenet service, easy to implement, and VoIP is a short step forward. IM is better than SMS, so it differentiates VoIP service offerings. Lots of potential synergy between Presence and VoIP (Next Gen Dialtone).
Six or Seven drafts in the works to define how IM can be done using SIP. Presence authorization (now known as watcherinfo). (One big issue in SIMPLE is over whether IM should be done with the Message method asynchronously, or whether to just set up a session end to end that can be used to exchange IM. Issues include trying to separate signaling and transport, and making sure it works through firewalls and NATS).
3 IETF groups proposing IM standards, but SIMPLE is going win because Microsoft and AOL/TW have endorsed it. "Presence spreads" -- once an application knows your presence for one purpose it becomes useful for others.
Tellme has an extensive PSTN infrastructure, but is looking now to interconnect using SIP with ITSPs. They have established many requirements to do this, and starting to get people to do it. Their experience included:
Tellme also migrated internally to SIP based telephony and commented on SIP migration requirements (for an enterprise):
Hybrid PBX and SIP -- The SIP phones connect to a gateway that look to the PBX like a remote PBX. All the components are available today. Decided to write their own voice mail system using their VoiceXML solution.
VON Panel Session.
This was on the early history of VoIP The common theme here was people involved Vocaltech and the recent history of VoIP. Too many panelists to keep track of who said what, so I'll just summarize some things people said about the industry.
Someone showed a chart on the foundations of VoIP, which went back to Voice on the Arpanet in 1973 (Comment, I remember this too. There was a lot of debate on whether or not this was crazy, but remember in those days the IMPs (The routers of the arpanet) were connected with 56K links for the most part).
Dan Berninger of Pulver.com said: "During the whole history Bell Laboratories didn't have anyone working on packet voice" He was a Bell Labs alumnus of some era. (Comment - actually this is wrong. Bell Labs research had an active program of packet voice on the ethernet during the late 1970's. In 1981 Bell Labs in Naperville launched a major program in packet voice for the PSTN known as Fast Packet Switching. It was influenced by the Arpanet work but really driven by integrated services and lowering voice transit cost starting from the center of the network. The program designed a scaleable hardware packet switch, gateways, network protocols, and techniques for flow control, quality management, and jitter control. The program was shelved in the mid 1980's after Sprint began advertising their all fiber network. This caused AT&T (and everyone else) to shelve many technology programs in favor of deploying more fiber. An additional issue with Fast Packet was that Sprint defined the voice market to be about best quality, and at the time, Packet voice was good, but at best only equal to PSTN quality.)
First IP phone call from VocalTech in 1993. First call was office to office. Someone thought it didn't work and went next door to check, but then the voice came out of his office -- 8 seconds delay!
One of the folks who then worked for IBM talked about concluding early on that: "Today there is Voice and Data, Voice is AT&T and Data is IBM, in 5 years there won't be any more voice." This was a pitch to IBM, IBM didn't get it, so the guy left to go to Vocaltech.
"Early on it wasn't about connecting phones at all" -- Vocaltech was focused on Voice over the internet and offered a PC VoIP client. Early on they piggybacked on the Internet Relay Chat system, but got kicked off of these servers because of concern for load. This really triggered the development of an independent registration server and independent protocols.
Industry goes through cycles. Early on it was PC to PC. Then everyone focused on gateways to do phone to phone. Now with Windows XP, focus will be back on the client.
David Gurle, now of Microsoft, was one of the panelists as an original member of Vocaltech. He talked about how he came to Microsoft, hired to basically do a voice client right. He said he was unhappy with the initial direction of Microsoft and threatened to leave, but was allowed to redirect the program more in the direction of open standards and interworking.
One question that was put to the audience was how many people felt that Voice over IP would dominate over the PSTN in the next 20 years. Interestingly enough few people raised their hands. There was some discussion on this, with the view being split between those who thought the PSTN would still be there much in the current form (Railroads didn't go away because people invented cars or airplanes) or would absorb VoIP technology and evolve into just another application on the global internet.
JAIN Update Session:
In spite of being at 8:15 in the morning, they had 20 or 30 people in the room for this. It ran up against update sessions for just about every programming interface and architecture (PAM, MSF, Parlay, Softswitch, SIP, etc.)
Margret Nielsen (Sun) -- Overviewed the JAIN stack and JAIN architecture. Nothing surprising here, but she has a good set of slides that go through the JAIN stack.
Jain SIP, JAIN JCC, and JAIN MGCP have just been released as final standard.
She talked about how to map JAIN onto different network architectures (SIP, PSTN, Softswitch, etc.)
She talked about the portability demonstration done at Supercom by NTT. Lucent working with Fujitsu was one of the participants in this, NEC was the other. The demonstration showed migration of code between different JCC/JCAT implementations.
She talked about the mapping of Parlay and JAIN, showing a JAIN/Parlay architecture in which the applications were written to JAIN SPA, which then essentially used Parlay as a secure remote procedure call interface to connect to the JAIN SPA level inside an application server or call controller in a service provider network.
She also had a chart showing the mapping of the OSA interfaces into Parlay and Jain. Good background.
JAIN announced a new certification process to certify implementations of the JAIN APIs.
Chris Harris (DynamicSoft) JAIN/SIP -- Actually focussed more generally on Java in SIP.
Why Java with SIP: Same culture of openness, lots of synergies.
3 Java APIS for SIP:
For focussing on the Media, the Java Media framework is probably the best set of APIs to use.
Ramana Dovarapali (Truetel) JAIN SLEE
I talked to him afterwards and learned he is a former contracter in the Lucent INU.
SLEE is a concept from the IN world. SCPs provide SLEEs to allow applications to be easily developed. The problem is they use proprietary APIs and tools and this limits the flexibility and makes portability impossible. JAIN SLEE offers the opportunity to provide the same level of support for a high level and secure development, but in a portable way in a widely support language and tools.
SLEE defines service lifecycle (install/remove/activate) Service component models, subscriber interfaces (subscription, management of subscriber data, etc.).
The SLEE defines the security model.
JSLEE defines an execution model based on events, service activities (groups of events), service instances (like a call). SLEEs can have multiple services and there is a model for interaction of different services in a SLEE.
The SLEE supports the lower level Jain APIs, provides the component model, and the framework for events and execution. It also provides a way to manage the services based on Java Beans.
The SLEE can plug in at any level. If you want to build directly on the low level APIs, you don't have to go through JCC. Currently the focus is on JSLEE as a service provider secure environment, but there is interest in using it for 3rd party application servers as well.
He had some examples, with a SLEE box (Application server), interacting with call controllers (softswitch). One was a simple routing where the SLEE was acting like an SCP doing a routing service. A second showed a call being routed to a media server, and then the SLEE interacting with the media server to interact with the call. Neither provided any real detail on how the SLEE specification worked. I would very much like to see that.
The speaker's company has a product which he billed as a JSLEE compatible application server.
Question: What's the status of certification:
Question: Is JAIN
SLEE the same as Parlay Gateway? His
claim was that JAIN SLEE is a more complete execution environment support than
Parlay, which is basically just an API. (Comment -- I'm not sure about this,
since the framework APIs for Parlay provide a lot of execution environment components. I guess what Parlay doesn't provide is the
actual event execution model.)
Question: Are there reference implementations? Yes.
Question: What about a service creation environment. Yes, there is a JAIN SCE group. There are 10 companies in it, they are looking at component approaches but also at XML as a way of doing it. The result may not be a traditional specification.
Question: Are there reference implementations for JAIN SIP Lite and SIP Servlets? Jain SIP Lite is too early, SIP Servlets is being developed and will be available to the java community very soon.
Question: Does the XML work overlap other XML based service creation? JAIN SCE is very aware of the other work being done on this and wants to take the XML work into the appropriate standards forum.
Session: North American Carrier Deployments
This one was crowded, probably 100 people. Its clear that when carriers talk, the VON community listens.
Traditional telephony is focussed on layer 1. IP telephony requires expertise through layer 3.
Verizon has had VoIP since March 2000. They provide connectivity from the public internet into each LATA, and then switch from there. There are 6 LATAs (Boston, NYC, Northern NJ, etc.).
Mostly Cisco equipment. Central NOC, with routers and gateways in each LATA.
For testing they have a separate test lab, and a full platform mockup to test equipment in a non live traffic network model.
Lessons:
Future:
Commercial deployment now is SIP based services for enterprise customers. This is different from most carriers going after class 4 replacement, tandems, internet offload, etc. This is really new services. Services include enterprise calling, on net. Off net via PSTN, private numbering plans, managed gateway services. Launched in March of 2001.
Worldcom has an internal deployment of this same architecture with 30+ corporate sites using the same SIP server (I think they built it). This has been in trial state for 18 months, and active for 6 months.
Upcoming (early 2002), SIP to the desktop -- centrex features, SIP Phone support, SIP aware firewalls, voicemail, and local calling.
Enhanced services (first half 2002) including multi party calls with bridging, presence and IM.
Foundation of their offering is SIP in the enterprise. In the enterprise there is a gateway from the PBX to the data network, Worldcom provides the data network and routing to interconnect sites. End users have PBX phones. Their network SIP servers that route the calls can access their PSTN based VPN offering databases for dialing plans so that they can transition from PSTN to SIP based interconnecting. There is a network gateway that offloads off net calls into the Worldcom PSTN.
Upcoming support just adds the IP phones. They also add local gateways that allow the enterprise SIP phones to connect locally to the PSTN for emergency services. In the network they add conferencing and messaging servers.
Enterprise cost savings is still the major driver for deployment. The driver varies a lot depending on things like:
Acheiving QoS: Need to look at 3 segments.
Results: 99.8+% of calls get less than 0.75% packet loss in real measurements. Lots of things to measure and manage here.
Protocols define the architecture, not vendors. She took issue with the statement by another speaker that carriers don't want to do integration. They must integrate in order to maintain multi vendor interoperability. This is why a standards based implementation is essential for them. They were very much involved in enforcing adherence to IETF standards and working out problems where the standards are ambiguous.
Interoperability issues:
E911 -- Problem is IP addresses don't identify location.
Worldcom wants to provide service components to a 3rd party ASP through a developer kit
Black phones won't go away. More Java/XML/Web programmers than IN Engineers. Everyone wants new ways of integrating multiple services.
They wanted to limit dependency on SS7 signaling and IN because they really feel that extending it to do multi media and support the web model wasn't productive.
They want to leverage the QWest assets -- CyberCenters, IP Backbone, dial access networks, and an IP enabled IXC network using Sonus equipment. They are currently replacing DMS 250's with Sonus gateways.
They announced last week a local switch replacement program with Nortel. They are carrying live traffic (ATM Voice) in Idaho.
Cisco is the support for their dial access service network.
Applications today:
He gave an example of sorting out the "good" and "bad" customers to prioritize them as well as trying to upsell to customers with old phones, poorly matched calling plans, etc.
This is carrier neutral, they can offer to customers using other carriers.
(Comment -- where was this guy when I was trying to pitch network IVR and call centers based on VoiceXML and application servers to some of the ILECs a couple of years ago)
Problem: QWest territory only includes 2 of the top 25 cities, so they need to be in other regions. Their answer is CyberVoice Direct, which is IP connectivity using either an IP carrier or ITSP model. They are in beta in California. Intially H.323 based with most available codecs. They are currently working on a SIP network in parallel. SIP will eventually allow them to originate and terminate traffic, though they currently have regulatory problems with this.
Question: Do you see H.323 and SIP coexisting forever of is there a transition. Answer -- Worldcom is SIP now, no H.323 now, will have to support it over time in an interworking model, like interworking with PSTN. QWest did H.323 because it's mature, the whole future is SIP based. Verizon agreed, though 323 won't go away. They want to have a 323/SIP combined platform.
Question: How do you provide survivability of services provided off of central servers to SIP phones on premisis if the IP pipe goes down? Worldcom -- First issue is engineering the network to make sure LAN and access networks don't go down. Backup strategy using internal gateway to PSTN. (Others nodded) Teressa's comment was she is more upset if the LAN is down than if the PBX is down today. IP is more survivable than PSTN because you aren't tied to a local switch. (Comment -- What does this do to the PSTN carrier's business model -- in effect they become the emergency backup. Can they provide that service at a reasonable price?)
Question: What's QWest's strategy for support of the enterprise? Looking at things like Worldcom, struggling with how much equipment to put on premesis. They don't want to do the local gateway solution to 911 because they are concerned about people who roam and are connected to the enterprise LAN via VPN.
Question: What's QWest doing about local access? Before US West acquisition they were planning a local access solution, they are continuing to do that.
Question: What's the schedule for higher quality voice over IP. Teressa -- It's clear that you can do high quality codecs, the deployment issue is when they become widely available so you can actually negotiate an end to end connection. Big cost issue is the access link (i.e. the problem of not being able to carry very many RTC sessions on a T1 alluded to yesterday). Getting a gigabit access network in place to remove the bottleneck is the solution.
Question: Bell Canada has an IP VPN service. This is a good platform to put more services. What do you have like this? Worldcom has a UUNet based product like that, they are about to launch a frame relay version of this. Tim -- Bell Canada is cheating because they are vertically integrated. If QWest wants to do this kind of thing out of region they run into tariff issues (i.e. if you want a fraction of a DS3 in Dallas you pay for the whole thing) Another is just scale "Not a whole lot of places to go in IP in Canada".
Question (Henry Sinnreich) Actually a statement about delay reiterating his keynote material that delay measurements are not very meaningful because it's very low most of the time, and extremely high during rare events
Tim Good points, but the real reason for VPNs and other techniques that control ingress into the network is security.
Question for QWest -- How much deployment of Voice over ATM and is there a migration to IP? -- ATM is only in a few of their local markets, it's a cost savings play and based on work that was initiated by US West. Tim sees a transition to VoIP beginning in 2003.
Question: (Medhi Gassem) -- Are capital limits limiting VoIP deployment? Teressa -- not at Worldcom, it's a different overlay business. They scale their IP business based on how much traffic they sign up on it. Timothy -- Has simplified their thinking because they have cut off a lot of things that they were pursuing. They basically have given up on doing any work to make black phones work with features out of the VoIP world. Verizon --- Slowing down a bit to look at the options before they spend, but they are planning on a major deployment.
Session: Network APIS:
Only 5-10 people in the Audience! I don't know whether this was timing or just what. I'd have thought that Doug Tait and Steve Davis would draw a better crowd.
Computing is a very old discipline, mathematics go way back, Networking is much more recent, brings a lot of new ideas to the pie. He talked about different perspectives. Networking people talk about protocols, computing people want APIs. (Comment -- I'm not sure I'd agree with the taxonomy, but I would agree that there are two ways to think about convergence).
The real focus needs to be on services -- then look at what pieces I need in computing and networks to determine what's needed.
Java Technology for Service Providers (JTSP) Sun had a well defined paradigm for the enterprise/data center, and environments for the devices. JTSP spans all of these into an end to end solution.
He then went through the Java stack, about the same as Margaret did in the earlier session.
Intersting timeline showing the evolution and adoption of Java, J2EE, and JTSP. Expect mass adoption of JTSP 2002 or 2003.
He went through a lot of the structure, again duplicating the presentation earlier. An intersting chart here showed the family of APIs and how Parlay, mobility, and others fit into the overal architecture.
He showed the SIP network architecture with SIP Servlets in the proxy and application servers, and SIPLite in endpoints.
Steve Davis (Ulticom) -- The real need is "show me the money." Parlay enables revenues, that's why it's being driven.
Lots of issues in wireless networks in the evolution from 2G to 3G. Enhanced services are needed to drive short term revenue.
Next Generation network trends -- no real advantage in converting PSTN to IP without new services. There hasn't been much innovation in new services yet.
Recent reduction in telco spending virtually guarantees that there will be mixtures of technologies in networks for the foreseeable future. This ups the importance of supporting hybrid networks and service portability.
Within 6 months of Java overtaking others as the most used most popular programming language. (Comment -- I'm surprised, I would have thought Visual Basic would still dominate, and I'd have thought the split with Microsoft would slow down Java).
"Programmable Network" -- based on concept of IP programmable switches pioneered by Summa IV and Excel (Comment -- interesting that these companies both have been consumed in acquistions, by Cisco and Lucent). Why not have the same concept with a whole network -- programmable network.
Parlay has been adopted by ETSI, JAIN, OMG, MSF, more to come. Parlay is language/distributed computing neutral. The Parlay group will provide an XML mapping of the APIs.
Parlay release 3 will be jointly released with ETSI (3GPP) very shortly. Parlay 3.1 in 1Q02, which is a cleanup only, plus the XML mapping. Parlay 4 is just starting, but there will be backward compatibility with Parlay 3.
They have an application server product TappS. Leverages a web architecture for service context management. The company is based in Israel, incorporated in the US (But currently no US personnel).
The quest for services -- we keep looking for killer applications in the telecom space, but there don't seem to be any. That's the wrong focus. Need to have a network that enables lots of services each focussed on a small community, not one killer ap for everyone. Group calendar, IM integration, etc.
What's out there? From their perspective Parlay is Corba focussed and langauge agnostic. Jain moves those towards the Java world. Implementations? Limited.
Moving forward means moving from standards to products. The industry forums have done a good job in definition, now we need products:
Reliable, high perfornace, "carrier grade" Adaption of existing protocols to Parlay/Jain.
Two types of products:
What's a telephony application server? IT/Application focused, not telephony focused. Managed framework that lets you access multiple parlay gateways from multiple vendors and integrate with other application servers. Non functional requirements (high availability, etc.) Used by both enterprise and service providers.
What's wrong: Too many standards bodies -- can't follow them all. JAIN/Parlay convergence is good but not enough.
No mature implementations.
Java APIs are more usable than Corba for broader community. (Comment -- I think this is true of the web community, I'm less sure about the enterprise IT community).
Billing/charging.
Other points:
Need to deal with the PSTN -> Lots of people there, services there, you need to intework.
Parlay/JAIN are difficult: Callback and event based programming is more difficult than typical business logic. (Interesting perspective, I'm not really qualified to address this one). Steve Davis got in to clarify that Parlay understands this and has created the Parlay X group focused on a simpler high level mapping using XML.
IT developers will have a hard time ramping up to use them.
Call control isn't the only barrier to new services: Database programming, Security, and other middleware (message queuing, transactions, etc.) are needed.
Java with ever increasing extensions is the best route -- Use Java.
Application servers will mask much of the underlying complexity.
Operators will install Parlay gateways only if it is economical.
As an application builder, build on a good application server. Still need developers with telephony background.
Where does Personetta fit? -- service layer, sitting on top of call control.
Question (me): How long before we get Parlay gateways or JAIN SLEE in big carrier networks?
Question (Brazilian company building softswitches) -- not clear what standards to go with. They have a softswitch based on an evolution of TDM technology. Doug Tait talked about the JAIN roadmap in response. Said JAIN JCC has been universally adopted by the softswitch vendors and that's the state of the art. Steve Davis -- industry doesn't buy technology, they buy revenue sources. Have to go with solutions that help do that. 3GPP, ETSI, and Parlay are in convergence, and in convergence with parts of JAIN. There is overlap between the rest of JAIN and the rest of Parlay. Nobody covers everything. If you want low level access, JAIN. If you want mobility, you probably need the 3GPP specific extensions. If you want the higher level functions, Parlay or JAIN SLEE, but they are working on converging these.
Doug also pointed out that a 3rd party or a softswtich/application server vendor can use the reference implementations of the Jain APIs and use them to kick the tires without having to spend a lot of money on acquiring products and setting them up.
Question: Java license has a disclaimer that it's not for use in safety critical systems. Isn't telecom safety critical? Doug -- clause is probably there to limit lawsuits. They are working on addressing these problems at JSLEE level. Not aware of any activities at the JVM and J2EE level. (Comment -- this is interesting, I never noticed the detail, It may limit liability, but this looks like a problem for SUN if people actually worry about it. I could see someone who built a java based softswitch which malfunctioned being held responsible because they ignored the warning.)
Question: Certification process, What about JTAPI? Doug -- Went through the certification process, said that JTAPI was a learning experience for how to do this kind of thing, JAIN will do better because of the requirement for the reference implementation and TCK (tests) for compliance. He also pointed out that you could use the same machinery to qualify Parlay implementations because they map.
Eric Sumner (Dynamicsoft) -- Jeff Dworkin (Pulver.com) summarized the message as: "SIP or get off the pot")
Explosive growth of SIP clients over the next 5 years driven by 3GPP and XP.
SIP is the winner, winning people in the application space will build on SIP, Losers will try to be protocol agnostic and build on protocol agnostic APIs.
XP and 3G are in spaces where people pay for services and money can be made. Plus there are lots of fundamental needs that aren't being met: 6/7 calls to business don't go to the intended recipient. Presence and IM are the tools to do it.
At Dynamicsoft presence changes the mode of communication. People first send an IM -- takes less time than dialing, doesn't get done until you see the person you want at their desk. Don't have to listen to ringing and go to voice mail. (Comment -- Yes, I hear this. In a previous life my office moved from forwarding to voice mail on 2 rings to forward on 4 rings, to save money on disk space, without considering the impact on the callers of the wasted time.)
Two of 3 "cartful" customers on e-commerce sites bail out before buying. Lots of reasons, lack of reinforcement and confusion. IM and Voice will help out this. (Comment -> Maybe this helps some people, and a year ago I was promoting an application like this, but I think you have to be careful how it's done. Personally I'd freak out if I was shopping and then contacted by an operator who seemed to have been looking over my shoulder. I recall a hacker on the MIT ITS timesharing system who did this to people playing games 25 years ago. It was really spooky to realize that someone was watching you and you didn't know it.)
SIP architecture -> Your service provider knows your preferences, billing and subscription. This connects to both the PSTN and dto XP and 3G clients. Service provider forwards to various application providers. Dynamicsoft has solved all the hard problems in service provider. The majority of carriers that have announced that they are going to terminate XP traffic are Dynamicsoft customers.
Wireless market: 1 Billion clients, turn over every 18 months. This is terrific because you get new features in quickly. Wireless carriers are great customers -- they are healthy and they know how to bill. Wireless subscribers pay an average of $50/month for services.
Both 3GPP and 3GPP2 are working closely with IETF and Dynamicsoft to insure that the standards will work and are correctly implemented.
What about 2.5G? 2.5G has a wide enough always on IP connection to do signaling in SIP and signaling for IM, presence and other setup functions. Example service:
Architecture looks just like XP (service provider and application server). One key thing is that the service provider will manage billing, subscriptions, all kinds of OA&M functions that make it easier for the application developer.
It's the services stupid. What's wrong with today? Today is expensive, slow, and boring.
Why be optimistic?
Key is following a 3 layer architecture with SIP/HTTP between the layers. Don't use proporietary glue!
Will see rapid development of innovative services over the next year, and will get services for millions of people.
Ike Elliot (Level 3) -- Cockroaches (and how to be like them) -- Our industry has been characterized as being in a Nuclear winter, and cockroaches are the only thing that survives a nuclear winter. (Comment -- what a change in mood from about a year and a half ago when Level3 was riding high!)
How did it happen? Some view from the Pundits:
Good news:Unlike the .coms, we have a market!
Bad news:Shakeout in progress and it will continue. The softswitch consortium lost 54 members this year (ouch!)
Good news: An explosion in progress -- Microsoft XP, wireless extension, cable extension. Softswitch consortium gained 35 new members this year. Mostly new startups.
Bad News:Tough times force a focus on established markets, danger is just recreating the old world.
Good news: Established customers are buying our products and services.
Who survives -- it's good to be a cockroach:
What's next:
Bob Schecter NMS (formerly Natural Microsystems) he is also an alumnus of Lotus.
The state of the industry:
Telecom in transition. Nuclear winter in the Telecom industry. Stocks trading from 5%-35% of 52 week high, 1.5 Trillion lost. This is 2-3 times the size of the .com losses (Comment – and that doesn't even reflect all of it, since Lucent was way down already 52 weeks ago).
Some sure signs why:
What's the problem:
Wireless is where the action is:
Some examples:
Voice applications: Phone meets web.
Early days -- voice portals using speech, IVR like, single vendor with unique content. Limited media and personalization.
All the major wireless providers have launched wireless web services. The web players have launched their own. Voice application developers (Tellme, Bevocal, Hey Anita, etc.) Big enterprise players (GM Onstar, UAL, Fidelity). Multi-modal devices. (PDA/Phone)
Tomorrow -- multi-provider content (standards based, integrated.)
Services layer -- $41 Billion by 2005, supported by $16 Billion in technology. (Comment -- big market for VON, not so big for Nortel, Lucent, Cisco, et al.)
What's needed: Network ready systems
How to get past Nuclear Winter:
1990's were the decade of the web, 2000's are the decade of IP Communications.
"This is the dumbest decade in history -- and we have 9-1/2 years left" -- Michael Dell.
What analysts are saying:
What happened to VoIP?
But International VoIP has clear momentum.
He had several charts of minutes. Very interesting. iBasis carries more traffic than lots of carriers, number 18 internationally. Most carriers only get about 5% of their revenue from international traffic, (interesting on his chart, Belgacom and Telmex were the big exceptions both over 15% international). Thus they don't see it as a core business and outsource to iBasis, ITXC, etc.
Interesting chart on investment in circuit and packet. On the chart it looks like we are about to cross over (more dollars into packet than circuit), but the scales are off by 10X. Circuit is still a lot larger. It would be interesting to look at the raw data and add up the total -- The total is almost certainly shrinking. The implication of that is that the dominant players in circuit are just plain up the creek. Even if the got ALL the packet business, it doesn't replace what is being lost on the circuit side. Much like comparable points in the mainframe->mini->PC transition.)
Domestic VoIP is growing slower. Particularly in North America, where everyone has a phone line. In countries with low teledensity VoIP may be a leapfrog technology and is being more actively pursued. Another factor is that while they used to fight technology to keep carrier jobs, governments are starting to realize that cheap telecommunications is a stimulant for growth, so encouraging competition and lowering costs isn't bad.
Another interesting chart on usage versus price per minute. Again it would be interesting to do some math and see whether total revenue is going up or down. I'll bet it again is down. Revenue from VoIP and services doesn't replace loss from the decline in price per minute.
Enterprise market. This is growing where people are willing to deploy. Not the same barriers. A major driver for this is simplifying mobility. You can't afford to keep paying for rearrangements. (Comment -> If this is true it is particularly important not to eliminate this savings by forcing some kind of centralized provisioning scheme in order to provide 911 service).
Jay Batson -- Pingtel -- Scenarios for technology companies in trying times.
Good news:
Bad news:
Questions people are asking:
Time to rexamine the business strategy:
Use scenario planning -- create a reasonable plausible but structurally different futures -- not a guess on what will happen but thoughts about what could happen. Not a prediction, a planning process to adjust business in response to risks and increase chance of success.
Pick a future point in time, create credible stories that describe that time, look at what that suggests and what the impact is on the general populace, take off your filters (time, proximity)
Examples: Shell oil
Here are some scenarios he proposed for VoIP:
ruinous savings:
prolonged recession causes lots of layoffs, scared consumers, falling markets, cash is king.
Some argument for this is the collapse of the baby boom. Spending goes down down in older age. You can predict GDP (and stock prices that track GDP) from from Spending by age and the age distribution of the population as projected from birth rates. It's actually pretty good except for the last few years, where the market got way ahead. The S&P is still 30% ahead of the projection -- not good news.
The networking business in this scenario:
Evolution wins over revolution:
What do we do?
(Comment -- cash out and retire :-)
Scenario 2 -- bullet dodged.
Europe's baby boom is 5 years late, and this helps kick start the world economy. End of recession is 2Q2002 New money enters the market to fund new business. US War against terrorism causes business boom.
Network market:
Impact on VoIP
Asian Surprise:
Networking impact:
Key is to be prepared for as many as you want (Comment -- Very interesting, Personally I hope there are some other scenarios in our future because none of these looks all that great.)
John Yoakum -- Nortel
Slides and Notes on pulver.com slides site.
Common characteristics of a busy person (soccer mom, executive, teens) –
He presented lots of stuff on presence and inferred presence by your location or use of communication and computing devices. (Comment -> Where have I heard this one before. Another great concept that has been slow to develop.)
Presence belongs in the internet -- where people go to find out about you. (But that may just be a pointer to something personal that actually holds your availability)
Looked at services integrating gaming and TV (Comment -- another one of those we worked on 1-2 years back). The idea was you could make a home with a TV and a Phone talk to an office with a multimedia broadband networked PC as a peer. He couldn't talk about all the scenarios because "people in North Carolina are busy filing patent applications"
He talked about using his wireless LAN laptop, with VoIP and a headset to basically do business from anywhere in the conference, and that the wireless LAN worked really well this time (Comment -- maybe because attendance was lower than planned?) This was a common phenomenon. Lots of people sitting in corners typing and talking on their PCs.
He went through a back of the envelope calculation on how many presence transactions there would be in a big deployment:
This drew a lot of flack from the audience over whether big centralized presence servers are "the internet way". The bottom line though wherever they are, presence servers are a big opportunity. (Comment -> Yes of course, I saw this one coming. The risk of course is that Microsoft will just provide this built into IIS, exchange or something else which will take care of a lot of this market.
It's not about Voice over IP, Voice over ATM, it's about Voice enabled IP services. (Comment -- Timothy is not a US West person, he is a true IP person so this is a very natural view.)
Convergence is happening, but maybe not the Voice model of how services work. Email and IM show different models of how services get deployed. Presence infrastructures are already there in a lot of applications that have no current way to share them. His example was your PC calendar service. (Comment -- advantage Microsoft again, since they probably dominate that market in the enterprise and whatever anyone does on the APIs for Outlook they can probably out-do by building it into a new release).
Users have some of the right tools, but aren't using them. (Complex services with poor user interfaces)
PSTN interoperability has been the only objective for VON. It can't be that, it's important, but it's only one of thing.
It's one thing to replace trunking, which is dumb, Quite another to replace an SCP. That will be harder. (Comment -- Interesting. My discussions with him before were always that he was anxious to get rid of SCPs.)
No OSS/BSS maturity in the IP telecom space. "QWest is a billing company -- we send 31 million bills." You have to demonstrate you know how to help us do that.
Security in VoIP is an afterthought -- it can't be.
Saving money isn't enough. He had a comparison of 3 different PBX scenarios. Yes, IP PBX beats digital (but not analog!), and hosted IP is cheaper, but it's not huge.
Reliability -- PSTN is designed so elements can't fail. Internet philosophy is to assume they do and engineer network survivability. with VoIP we often demand both, which makes for huge costs. Can't do that, should move rapidly to the IP model.
The concept of Long Distance is going away. Distance doesn't matter in IP. The concept of "Central Office" is also going away -- You won't have a single place on which you depend for your services.
Have to do a better job with identities. 8 phone numbers for one person isn't useful. At the same time we have to be very careful to avoid an identity theft problem.
How to sell to QWest:
He closed by asking the audience who Strowger was. I knew this was a "different" audience when after about 15 seconds I answered it. His message was to remember Strowger for several reasons. He really invented the circuit switched network that has persisted for 110 years! He solved a business problem. VON has to do the same. If we can't do better than an 1890's undertaker, something is wrong!
Accelerating VoIP deployment needs good voice QoS. Today, that requires a dedicated network where quality can be controlled. A dedicated network solution isn't acceptable from a cost perspective.
You can solve the problem of packet loss with overprovisioning in the core network because the bandwidth is there and it's cheap, but it's not cheap in access. Typical approaches (MPLS, call admittance, etc.) are band aids and move away from packet network back towards circuit.
While data is dominant over voice in bits, Voice sends much more packets and will dominate in the number of packets for some time -- you can't afford special voice packet processing.
The solution is to look at special packaging for voice, not special handling, like egg cartons that protect fragile eggs from less than careful handling.
Sound packets are fragile in that time matters -- you can't retransmit.
The usual problem is a lost or late packet is gap in the speech, lost meaning. Perceived quality goes down very quickly with packet loss.
Their solution is to repackage the speech so that if a single packet is lost, the resut is "blurry", not completely lost. The technique is like GIF coding, where a small fraction of the image data produces a blurry image that gets clearer as more of the data is recieved correctly. This is done by holding a few packets and spreading the speech across several packets, as well as using a coding that allows something crude to be reconstructed from partial reception.
In a followup in their booth, I learned that there are actually two parts of this, one as described above is end to end and works only if both sender and reciever use their coding and packetization technique. They also have a technique that improves jitter and lost packet performance when implemented only on the recieiving end. this works mainly by very quickly adjusting the amount of buffered speech and using a combination of interpolation and very quick recovery from a late packet to reduce buffering time while reducing impact of packet loss. This can be done with any codec on the sending end.
The result is to allow packet voice to coexist with data and survive with very large packet loss rates without significant quality loss.
He had several curves of "Mean Opinion Score" -- a measure of voice quality, versus packet loss demonstrating that both the single end and double end techniques they used made a significant difference.
(Comment -- parts of this are similar to what was done in the Fast Packet Network work done by AT&T 20 years ago. In that network, voice packets were laid out in a way that they could be shortened by dropping bits under network congestion, and the result would drop the least significant bits from the voice samples, essentially resulting in smooth degredation rather than a sharp quality loss. That network also had a technique to measure the "variable delay" experienced by packets in the network, which reduced the need for long "jitter buffers". Neither is particularly applicable to a standard IP network today. Global IP sound has a very interesting and valuable approach, if they can protect the Intellectual property while not discouraging broad deployment of the technology).
Question: How much delay do your techniques introduce? Answer -- depends on how much packet spreading you do, but the experience is the shortening of the jitter buffer more than makes up for any increased packetization delay on the sending end.
Question: Does this work with speech recognition? Answer -- yes, we've tested it, but packet loss does degrade recognition (no details on how much)
Question: The internet has essentially 0 packet loss most of the time and rare "burst" losses. Have you modeled it like that? Answer -- while the essentially 0 loss is true in the core, it's not true on access where bandwidth and proxy limits cause congestion. Their model for packet loss is markov, with a higher probability of packet loss following a lost packet, which results in a distribution with bursts of losses. If you lose enough packets in a row of course nothing can really help.
Question: How do you make your codec universal? Licensing costs can be a killer. Answer first of all VoIP protocols have codec negotiation which is the way you allow different ends to negotiate a common protocol. His view on how to solve the problem of deployment was for XP users to be able to buy the codec for a one time retail cost of about $10 (probably not excessive.) They are also pursuing licensing it to the terminal manufacturers.) He felt that in Codecs, "standards" are less important than "footprint" Let the market decide. (Comment -- again in followup I asked about the viability of users buying and installing CODEC's for XP given that Microsoft said currently the CODEC APIs in their client weren't open. He said yes, the real plan is to sell it through Microsoft. He also said that all of the major IP phone makers had already signed up to license the technology.)
Niel Ransom (Alcatel)
(Comment -- many years ago Niel and I worked together on various projects in Bell Labs including the Fast Packet Network described elsewhere here.)
Neil gave a straightforward overview of the VOIP industry, most of which I did not take notes on because it's well understood. Some interesting comments included:
Why redo voice?
At this point in time VoIP networks may actually cost more. (Comment -- very interesting. 2 years ago everyone was talking about how cheap VoIP was compared to circuit switching because of the cost of the circuit switches compared to routers. This was a bogus comparison, but people adamently defended it. At this VON, I didn't hear anyone claiming that VoIP was good because it was really cheap and several made the point that a fair comparison was likely to result in it being more expensive in the short term.)
What's behind the Telecom Slowdown?
Arun Sobtri (IP Unity)
Megatrends:
Status of VON:
The real
issue here is VoIP is real
He showed
a chart of the VoIP equipment market. Interesting enough the whole thing looked
to be about $500M. As observed
elsewhere, this looks great to a startup, but its miniscule compared to the
market for circuit equipment. We
haven't crossed over yet.
Next
Generation Network will be many years in happening. The price of core services
(long distance) will continue to drop rapidly
He put up
a chart mapping rate of adoption of IP and rate of adoption of services,
dividing the space into 4 quadrants:
He gave
some statistics on phone sales and commented that when he was at Motorola, Cell
phone sales hit a wall, slow growth.
Nobody saw it. Meanwhile they
also looked at Pagers. Nobody in the US
thought Pagers were interesting but they boomed in China, in spite of China
having few phones so who would use a pager, since they send phone numbers? The answer was the Chinese used them at
first to communicate lottery numbers.
Once deployed they found other uses.
One was sending messages coded in numbers, kind of like SMS using short
codes for messages.
Venture
capital is hard to get. All VC's want
to see a positive second derivitave of revenue/earnings (Growth accelleration.) If you can't do that, you don't get money.
Challenges
for the industry:
There are
lots of service out there, but no killer apps.
VON is
about Decentralization of the PSTN.
Voice traffic, control and the local loop are being unbundled. 9/11/2001 was a wakeup call for people to
recognize that their service depends on a single point, the CO. (Comment -- interesting, about 15 years
ago now there was a major fire in an office in Hinsdale Illinois that similarly
knocked out lots of services, because everything in the western suburbs of
Chicago was hubbed through that office.
The response at the time included a lot of businesses demanding dual
homing, and even some public calls for better network design. It costs though, and either the concern
never spread to New York or people's memories fade over time).
The
internet is all about multi-homing -- network reliability not element
reliability. (Comment -- again, watch
out for the physical connectivity. Some
of the people who lost service in the Hinsdale fire thought they were dual
hubbed logically, but physically all routes went through the same office)
There are
3 options for business services
Technology
makes the hosted model possible and attractive -- like web hosting or personal
services through a portal site. You
don't care where the host is.
Why aren't
we moving faster?
===>
Because we lack a simple common vision.
Something like "put a man on the moon by the end of the
decade".
"universal
service" has been the common vision/goal of the communication industry,
but he put up a definition of it from the 1996 telecom act and it's not very
exciting or inspiring.
Who sets
objectives?
Message is
that we have lost our common vision.
We need a
national communication policy providing a clear vision. He laid out some he would like to see:
(Comment. This was about the third speaker calling for
what a year ago would have been heresy -- government establishment of goals,
policy, etc. Amazing. He is aboslutely right though about the lack
of vision. I am surprised though that
he didn't give VON and Jeff Pulver some credit for establishing a VON vision if
not a universally agreed on one.)
Circuit
won't go away quickly, but investment in services will be limited, especially
in international gateways, which is where they are.
Real
driver for VoIP is services, and especially convergence. People want multimedia.
Roll out
of DSL (Current North American problem is a temporary glitch). 9/11 will push this faster since corporations
want their employees to be able to work from home.
Internet
traffic continues growth (100% CAGR internationally) but revenue isn't growing
much because of falling price per bit.
Network efficiency has to be tight.
Voice
minutes grow -- all of it in VoIP.
Disinvestment
of vendors in circuit switch enhancements will cause maintenance costs to rise
as business falls.
He then
went into a whole bunch on Teleglobe's business, a competitive provider of
international long distance and IP service on all the major cable routes, and
providing a lot of IP services (hosting, VPN, POPs, etc.) Feels there are a lot of communication
applications that can be hosted out of this kind of infrastructure. They are running a softswitch trial. Nothing revolutionary, but they are really
doing it.
Question: Hosting, how many sites? Answer -- could do it from one if they
wanted to, cashing allows for quick response everywhere. They do provide multiple sites, but it's a political
decision, making people feel comfortable about having a hosting site on the same
continent. (Comment -- This will
work fine for content, but I really wonder about control. How many message exchanges are there between
a softswitch and gateways and media servers?
What would adding 20 milliseconds to each interaction do to performance.
Opher
Kahane (Kagoor)
Then: Capital is free,
debt is good
Now: Cash is king.
Then: get "Telecom eyeballs" -- no matter what the cost
Now:
It's all about the bottom line.
Then: if you build it, they will come.
Now: Prove it
Then: Experiment with
interesting services on the chance that something good happens.
Now: ROI, bottom line.
Have to squeeze
more traffic onto networks to improve margins.
Now, IP carriers overflow to PSTN.
This is really expensive, negative margin. Instead, we have to peer with eachother and if necessary use the
public internet.
VoIP has
something like 30-50X overhead (Goldman Saks) because of layering and QoS
issues). Have to do a lot better than
this!
One of the
big problems with Qos is that the IP Telephony signaling model doesn't help
you. Signaling goes to proxies and
softswitches but these elements can't enforce policies on the real
traffic. Need something in the routing
network that can enforce the QoS limits and policies.
Solution
is basically analogous to the web load balancing and content delivery
boxes. VoIP needs a "middle
box" that can enforce QoS issues.
(Comment -- like the IP Services Switch product Lucent/Springtide and
now many others build.)
The Voice
delivery problem might wind up being outsourced, like Aktomi handles media
intensive web services for various people.)
Who is
Tier 1? -- big incumbants in North America. (Comment -- I'm surprised he didn't
list anyone outside of
North
America.)
Why are
they important? They get an enormous
share of revenue and overwhelming share of profit -- they can afford to
buy! Intersting point on one of his
charts was revenue. Verizon and
AT&T were quite close in total revenue.
Local
service -- most of the lines are residential, most of the revenue and an even
higher fraction of the profit are from business.
Line
deployment turning down -- 2Q01 was the first in which the total number of
lines shrank. Total revenue dropped as well.
Wireless
is still growing pretty well, Cable is growing, but only slowly (I think this
was cable households, not broadband on cable).
RBOCS have 96% of the residential lines and 92% overall. (i.e. CLECS and independents are insignificant.)
Who sets
policy in RBOCS?
How does a
new player break in: By taking
advantage of disruptive technology to display disruptive economics in increasing
profit and reducing expense, but still nees a compelling need.
Telica
just got a Verizon contract (Comment.
This is very confusing. Some
characterized this contract as being "Class 5 switches", while others
have told me that Verizon has another RFP out on that still pending). The Telica people told me that the Verizon
deal is about PRI/data offload for now, with the potential to grow as a tandem replacement
or a class 5.)
Question
(Carl Ford) -- How/Where are the savings in operations, do you connect to
OSMINE? Answer -- Yes, they use OSMINE because
RBOCS require it but it needs extensions for packet, and doing that right is
needed to get savings.
Question
(Carl Ford) -- What part of the solution to Telica provide? -- Answer -- any part you want. Telica has all the pieces but uses open
standards and is testing with partners to provide different pieces. Their expierience though is carriers want to
buy the whole solution from one place.
Question
-- what are drivers for IXCs -- Answer -- mainly the same as LECS. IXCs are getting into local, and want Class
5 too (Comment -- this may be a better opportunity for a "Class 5"
softswitch, since the IXC's local entry will be targetted to busines, like most
of the CLECs were, meaning simpler feature sets in general and fewer issues
with loops).
Question
-- how are you doing internet offload -- where do you intercept the
traffic. Answer -- some at the trunk side of the local
switch, some at the tandem, some after the Tandem.
In spite of
the economy and threats of terrorism, VON had a good set of exhibitors,
probably around 100 companies. What was
notable to me though about the exhibits was there wasn't much notable. Nobody
dominated the show and nobody announced a new product that blew away the
competition. The big companies (Lucent,
Nortel, Cisco, Alcatel, etc.) were either not there or had only a modest presence. The biggest and guadiest booths were mainly
from component and software vendors (Microsoft, Radisys, Audiocodes, Dynamicsoft),
rather than the softswitch folks.
One
notable trend is that most of the vendor's of softswitches and application
servers are now promoting their product as doing everything. It's a gateway, It's a softswitch, It's an
application server, It's a service node.
Perhaps this is a symptom of what I have called "Famine on the
Seringetti -- in a healthy ecosystem, the animals have adapted to use different
food sources and avoid direct competition.
When some of those food sources dry up however, instead of simply dying,
the animals that were dependent on them start eating everything they can find,
and everyone runs short of food.
Companies competing in a bad market behave a lot like this, expanding
their product and their claims to try to capture markets they weren't in
before, driving up competition creating sales pressure for everyone.
The
Compact PCI based Sun processors were showing up in a lot of product
demonstrations, though one of the sun representatives told me there was concern
that these had been overspecialized for the applications in wireless base
stations and other deeply embedded systems.
GNP Computers is still there, though no longer in the softswitch
business. Bruce Rostofsky (Formerly
4ESS) showed their failover technology doing a failover and hot swap in about a
second and a half. It depends on
software checkpointing to achieve that level of performance. They were also demonstrating it on ARM processors
running Linux, mounted in little 2 inch cubes.
They said that many people are now looking at distributing all the control
functions into these kinds of processors sitting on interface cards and looking
for good failover solutions.
Of the
softswitch/server crowd it was hard to identify any real standouts. Most are trying to do everything. Most have at least trial deployments, but
few have lots of customers yet. Some familiar
players (e.g. iPverse) either weren't there or were there only with a person in
a booth.
The
protocol players were all touting SIP, as would be expected at VON. In addition there were several companies
exhibiting speech coding and buffering technology, probably the most notable of
which was Global IP Sound, with a broadband codec and QoS improvement software
described elsewhere.
There were
two interoperability labs on the floor, one dedicated to those who had
successfully demonstrated at the SIP show (SIPOP), and one open. Both were low key testbenches with a dozen
or so participants testing against each other and occasionally demonstraging
for people passing. I spoke with the
Lucent Technologies folks in one of these areas who had supplied a gateway and
SIP server for interconnection with the PSTN and interoperability was good,
though they had trouble with their PSTN connections because they had planned
coordination with several companies to provide an SS7 link and not everyone had
shown up yet.
Microsoft's
booth was much like at other recent events, demonstrating the XP SIP client and
APIs for both clients and servers. They
got lots of traffic, as did Dynamicsoft, demonstrating their SIP server. Several companies identified themselves as
Dynamicsoft partners in delivering a solution.
In one discussion with Dynamicsoft people, they claim that DynamicSoft is
providing the SIP servers for most of Microsoft's partners in providing Windows
XP Messenger based telephony.
Perhaps
one reason things seemed a bit subdued is it was clear that a lot of companies
are struggling. One exhibitor told me
that a dozen companies had cancelled at the last minute because of financial
problems, and 20-30 more would probably withdraw from the market or be acquired
after 3rd quarter earnings were out.
Not unexpected in this climate, but a bit of a downer compared to recent
VON events.