Where
do you put your Services?
An
essay by Warren Montgomery (wamontgomery@ieee.org)
Innovated communication services are essential both to
carriers and to end customers. For
carriers, services are what differentiates them from competitors and allows them
to retain customers and charge a premium rate for what is rapidly becoming
commodity transport. For end customers,
services are what allow them to increase business and personal productivity and
get the best value for their communication spending. In the past, most communication services had to be implemented
inside the switches and other network elements in the carrier’s network, which
was difficult, slow, and expensive.
Rapid change in communication and computer technology has made it
possible to position services in many different places now, and the choice of
where services are positioned has a profound effect on flexibility, security,
and cost for communication services
1.
Where do services go?
. Figure 1
illustrates some of the options:
- In
the user endpoints (phones, PC’s, private phone systems, etc.)
- In
the core network components (Switches, softswitches, routers, etc.)
- In
network based application servers (Intelligent Network SCPs,
web/application servers, etc.)
- In
network based media service nodes (Service nodes, media servers, etc.

1.1.
Endpoints
Many people looking at the internet immediately suggest the
only right place for services is in individual user endpoints.
- Because
end users control their own endpoints, they can install new services there
without without waiting for carriers to act.
- Entrepreneurs
can market new services directly to end users, creating a mass market for
innovation like that of PCs and PC applications.
- Many
of the security, reliability, and privacy issues associated with network
based service implementations are not a factor in the endpoints.
- Endpoints
are replaced more often, creating more opportunities for innovation
Not all services will work in the endpoints, however, and the limitations of endpoint based
services are critical to understand
- Many
users use more than one endpoint.
Implementing services in each endpoint leads to inconsistency and
duplication of data such as personal directories and profiles.
- Many
endpoints are not always connected to the network, meaning that any
service that needs to be available at all times, such as call coverage,
can be lost if it is implemented in an endpoint.
- Services
that require multiple media streams, such as conferencing, may be very
inefficient to implement in an endpoint, as it may not be cost effective
or even possible to deliver all the information.
- Privacy
and security concerns make some services difficult, because the users
associated with the endpoints may not trust each other enough to permit
sensitive information (e.g. caller’s identification) required for a
service to be delivered.
Endpoint based services will have a large role, especially
for convenience features (speed calling, local directories, distinctive
alerting, etc.) but will not be the only place fo innovation.
1.2.
Core Network Elements
Embedding services in core network elements, such as
switches, gateways, routers, etc. Has been much maligned as reproducing the
“mainframe” era for communications.
These network elements are nearly always under severe performance and
reliability pressure, meaning that innovation here is difficult, and lead times
to develop, install, and test new services are long. Still there are many advantages to embedding services in core
network elements:
- Performance
– This is the only acceptable place for some services, especially those
that process complete media streams,
or monitor every communication through the network.
- Security
– Services that must be provided, such as legally mandated wiretapping,
emergency access, or even collecting critical data for billing and network
management are natural candidates for core network services.
- Universality
– Services that must be provided consistently the same way everywhere,
such interpretation of naming, numbering, and routing plans, are good
candidates for the core network.
Still, in the past the communication industry has generally
erred in the direction of putting too many, not too few services inside the
core network. When the service does not
naturally go there this leads to many problems.
- Limited
capacity – Core network elements often have limited processing capacity
and severe performance requirements.
This makes it difficult to add new services.
- Impact
on security and reliability – Failures of any service implementation in a
core network element can impact many customers, so this is no the place
for rapid introduction of new services with limited testing.
- Long
lifecycles. Core network elements
typically have long lifetimes, making it difficult to incorporate the latest
technologies in them universally.
- High
specialized hardware and software – Meaning that few engineers have the
skills to work with them and service development can be difficult.
- Limited
numbers and markets. The total
number of switches in the US telephone network is in the thousands, not
the millions, like PCs. This means
the cost of developing applications for them must be recovered over a
relatively small market. Routers
and Softswitches face similar challenges.
In short, the core network elements are good places for
services with extreme needs for performance, reliability, universal
availability, or legally mandated availability, but a poor place for many
others.
1.3.
Network Based Servers
Network based servers are servers such as Intelligent
Network Service Control Points or Internet Application servers which are
located and managed inside of a communication network, but not in the flow of
every communication. The key difference
between these and the core network elements above is that user communication
requiring a service provided by a network server are routed to it, while other
communication does not. This gives
Network based Servers many of the advantages of core network elements without
many of the disadvantages:
- They
can provide a trusted intermediary between users and provide services
requiring information that the users would not trust to another endpoint.
- They
can be always on and available, providing services for users whose
endpoints are not currently connected (essentially as a proxy for a
user)
- They
can serve a variety of different security and reliability needs. Because users are connected to network
servers only when they request a specific service, the security model is
like that of web services – Errors in a website only impact users who go
there for services, not other services and other users. This enables more rapid innovation and
a more open environment for introducing new services.
- They
can be installed quickly and scaled to provide performance.
- They
can provide the same service for all of a user’s endpoints. A corporate directory in a network
server is available to a user on all his or her endpoints.
- They
can provide services efficiently.
Routing for media streams can be optimized inside the network,
avoiding inefficient “hairpinning”
Not every service belongs in a network server, for many
reasons.
- Network
servers often have limited access to control of communications and no
access to actual media streams, limiting the services that can be provided
to those supported by the enabling protocols used to communicate with the
core network elements. Intelligent
Networking, SIP, H.323, and other communication interfaces all provide for
network servers with this kind of limited access, but each have
limitations that make some services
difficult to provide.
- Coordination
between network servers, endpoints, and core network elements can lead to
service interaction issues. The
web is a good example. Each
e-commerce website establishes it’s own user interface and user
profile. Users often must enter
redundant information, and implementers cannot necessarily make use of
other services.
- An
open networking model can make it difficult to guarantee service
availability and performance. With
the telephone network, users have become accustom to a network operator
being able to guarantee a full suite of services. With services implemented in
application servers owned by different providers, no single party may be
able to do this. Regulation and
legislation may need to catch up with implementation.
I believe that Network based Servers represent an important
opportunity, and the implementation best suited to a large number of innovative
services. Getting the protocols and
interfaces right, and making the regulation and legislation fit this implementation
model is critical.
1.4.
Media Servers and Service Nodes
I make a distinction between an application server, which
deals with signaling and control for establishing communication, and a media
server or a service node, which deals with the actual communication
streams. Some analysts lump all of
these under the term application server, but I feel the distinction is
important because application servers as I have defined them can provide many
useful services without needing to deal with the capacity and interface issues
created by intervening in the communication media. In general a media server processes only the media under
direction from another network element (core network element such as a
softswitch, or a network server such as an application server or SCP), while a
service node is self processes both
media and signaling associated with communication. Service nodes look to the network and end users like just another
user, while media servers operate like a part of the network
infrastructure. They offer advantages
in implementing services
- Both
provide a good platform for media intensive services that are not needed
in every communication. These
include translation, storage, and many others. Many of the interesting service opportunities, such as
unified messaging, music and video rental and purchase, advertising, etc.
require media processing and are only really effectively implemented using
media servers or service nodes.
- Service
nodes provide a way to rapidly introduce new services without requiring
new network infrastructure.
Because they look like any other endpoint, they can be deployed
rapidly to meet a new opportunity without updating network elements to
either install a service on them or insure calls are routed to an
appropriate application server.
- Media
servers are provide high capacity and performance. An endpoint can implement some media
based services, but a true media server does it more cost effectively at a
large scale.
Media servers and service nodes do, however, have many limitations.
- Hairpinning. Because a service node looks like an
endpoint, it may be difficult to get efficient routing for services that
use them. This ties up extra
interfaces on the server and the network elements it connects to as well
as sending additional information through the network. The cost of bandwidth may be low, but
the cost of interfaces often isn’t.
This is a particular problem for networks and services in transition,
where different technologies are used.
Making services such as card based calling efficient in a circuit
network with established standards and a limited number of vendors and
operators has been difficult.
Don’t expect them to be efficient in a hybrid circuit/packet
network with much more fluid standards and more operators and vendors.
- Interface
density. Bringing lots of
bandwidth into a server is always a challenge. This is especially true in circuit networks, but is an issue
everywhere.
- Scaling
and operational efficiency.
Service nodes and media servers are great places to trial services,
but as the trial service becomes universal the number of servers required
and consistently administering them often gets to be burdensome. Many network services such as calling
card, voice mail, and others have gone through an evolution, starting out
as provided via service nodes but reaching a point where the number of
service nodes and the complexity of administrating the network drove
operators to seek other alternatives.
Media servers and service nodes are an essential part of any
network strategy, but it is important to keep the limitations in mind.
2.
So what are the implications?
All of these locations are places for innovative services,
the key is really understanding the relative strengths and limitations each
offers, and understanding how to use these elements together in innovative
ways. Some implications for carriers,
equipment manufacturers, and service innovators include:
- Insuring
interface and signaling standards will support the requirements of service
implementations. The key standards
include but are not limited to Intelligent Network, PINT/Spirits, SIP and
H.323 proxy services, Megaco /H.248 for gateway and media server control,
VoiceXML, and emerging standards for presence.
- For
carriers, deploying the latest interface and signaling standards to
support open network models for network based services.
- Network
based servers (Application servers, SCPs, and others) in my view represent
a real opportunity, provided the required interfaces are deployed. They offer many advantages over
endpoint or core network element based services.
- For
everyone, be aware of the strengths and limitations of different
alternatives, and be aware that the real issue is often coordinating
multiple platforms to deliver a valuable service.