IP Call Control Still Under Construction
By Charlotte Wolter
Building the standards for call control in IP
networks is an ongoing process, though the effort has reached an important milestone with
Large-scale deployments of Internet protocol (IP) telephone networks still have failed
to materialize, despite several years of feverish interest and investigation by service
providers. The holdup? The infrastructure of industry protocols and standards remains a
work in progress, and there is no end in sight.
The problem has been particularly acute in call control. Protocols have been
incompatible or don’t offer the capabilities of traditional networks. Network operators
don’t want to switch to an IP infrastructure if they have to tell their customers that the
network can’t deliver the services they are used to, such as call park, call forwarding
and toll-free calls.
Jim Nelson, co-CEO of protocol developer dynamicsoft, West Orange, N.J., says a number
of large IP telephone projects have been delayed by lack of compatibility and scalability.
"We need to move the industry out of trial mode to hundreds of thousands of
ports," he says. "Until we have interoperability, we will not be able to do
Lack of standards is also an impediment to service providers that want to move out of
simple toll bypass to enhanced applications, such as click-to-dial and web-enabled call
London-based British Telecommunication plc’s "only real option" for its IP
telephony venture in Spain was to build a single-vendor network (with Nortel Networks,
Richardson, Texas), says Bryce Kelly, signaling and middleware design, BT Advanced
Communications Technology Centre. "There was no standard, so no way to procure
against a particular standard, and everyone had their own implementation."
Some of these issues finally may be addressed as the International Telecommunications
Union (ITU) and the Internet Engineering Task Force (IETF), in an unprecedented show of
cooperation, are moving toward the adoption of a key new standard call-control protocol by
the beginning of 2000.
Called, in stolid ITU fashion, H.248, it is a child of the original media gateway
control protocol (MGCP) that the IETF touted last year, and is now even more tailored to
the needs of large networks. It provides a standard way for media gateway controllers to
communicate to media gateways.
H.248 is a critical standard, Kelly says, "because a significantly high proportion
of the costs of an IP network are in the gateway layer. So it’s really quite important to
be able to competitively procure that layer through open interfaces. Very important."
H.248 is an important building block in the development of next-generation telephone
networks, but not the only one needed. Network operators may still have to contend with
proprietary technology in their new networks, particularly if they want to implement all
the features their customers have become used to in the old networks.
"The most significant thing going on is the separation of call control from the
gateway, which is what the [H.248] standard is all about," says Morgan Littlewood,
director of business development, Cisco Systems Inc., San Jose, Calif., and president of
the Multiservice Switching Forum, an industry group for services development. The other
significant area of development, Littlewood says, is effort to unbundle services from call
control, so third parties can develop services for all platforms by writing to universal
advanced programming interfaces (APIs) and other interfaces, such as Palo Alto,
Calif.-based Sun Microsystems Inc.’s Java in Advanced Intelligent Networks (JAIN) effort.
Early on, the ITU promoted H.323 as a standard protocol stack for IP networks. H.323
had been developed originally for PC-to-PC conferencing. It was revised–the group is now
on version 3–and tweaked to handle telephony, and has been adopted widely by vendors and
network operators alike. These same users, however, have been vocal about its limitations,
and large network operators–such as MCI WorldCom Inc. and Level 3 Communications Inc.,
Omaha, Neb.– loudly have proclaimed that H.323 cannot scale to control a huge public
"H.323 was designed with an Intel [Corp.] terminal in mind," says Chandresh
Ruparel, product manager, IP telephony products, Dialogic Corp., Parsippany, N.J. "In
the public network, though, we have large switches, and the functions that switches offer
today are not addressed by H.323." Later versions of H.323 are trying to add those
functions, "but progress is slow," he adds.
The new H.248 is not a rival to H.323. H.323 is a way for media gateway controllers
(MGCs), the "brains" of IP phone networks, to communicate with each other and
control the basic functions of a telephone network, such as establishing calls. H.323 has
a rival in a protocol developed by the IETF, called session initiation protocol (SIP),
supposedly a more simple, direct and standard way to control basic network functions.
Each has a slightly different emphasis. "H.323 is addressing the various gateways
to other networks and other legacy systems," says Douglas Tait, intelligent network
(IN) specialist, Sun Microsystems Inc., Mountain View, Calif., and head of the protocol
expert group for Sun’s Java-based application interface, JAIN. "SIP is focussed on
setting up sessions and keeping things flowing between end points."
SIP receives much favorable comment in industry forums, but has only very limited
deployment. In the United States, MCI WorldCom is the most significant adopter of SIP in
its IP networks, though Cisco is said to be ready to adopt SIP in its platforms. Part of
that may be because of longevity: H.323 has been in wide use for more than three years,
while SIP is less than a year old.
Also, it would not be too difficult to translate between SIP and H.323. SIP could gain
more adherents as more is demanded of IP telephone networks.
"H.323 will not cut it at the edge of the network, because of the speed and the
[large] size of the code," says Chris Berluti, director of business development,
Iperia Inc., Wakefield, Mass., developer of a software platform for delivery of enhanced
services in IP networks. "With the high growth at the edge of the network and
next-generation POP (point-of-presence) equipment, there is work in standards groups to
make that [equipment] interwork, not just with the MGCP and SIP layer, but also with layer
3 quality of service (QoS), such as multiprotocol label switching (MPLS), and that will
leave H.323 behind."
If H.323 does not deliver functions such as call park, hold and transfer soon, "at
some point service providers are going to say, ‘We won’t wait any more and will go with
SIP,’ just as MCI did," Ruparel says. If SIP continues to be seen as fast and simple,
while H.323 is considered slower and more cumbersome–though it does more to support
multimedia–SIP could become the protocol of choice for vendors of large-scale telephone
H.248 is a way for the MGCs, the brains, to communicate with media gateways, the basic
workhorses of IP networks. One MGC can control multiple media gateways. It assigns
addresses and priorities, and instructs the gateway to stream out the packets for each
call. H.248 can work with both H.323 networks and SIP networks. It is also more
hierarchical, with more centralized control, like a traditional telephone network.
As important is the fact that H.248 enables a network where gateways are spread over
many remote locations and controlled by one box, the "decomposed network" that
many see as the IP-based architecture of the future, at least for the network edge. Nelson
says, "H.323 has done gateway to gatekeeper media streaming and call control tightly
coupled. That did not create an environment that made scaling easy. But with the [H.248]
effort with SIP, both out of the IETF, it promotes an environment that can scale, in other
words where the signaling and medium are separate."
After H.248 will come efforts such as the Multiservice Switching Forum, a group of more
than 50 switch vendors and service providers that is defining a next-generation switch
platform for the network. "The goal is to come up with an architecture that will
support traditional telephony and next-generation multimedia IP services and use things
such as MGCP, SS7 and ATM (asynchronous transfer mode) and a new switch control fabric,
all of which leaves H.323 way behind, at least in the edge of the network," Berluti
"From the Alcatel point of view, we have defined an initial architecture to
provide intelligent network capability," says Syed Rahman, product line manager,
Alcatel Assured Access, Milpitas, Calif. Although, "the ultimate goal is to have
every component within a network be able to work with any other component, we are going to
see some proprietary implementations for the short term."
With this new standard in place, will there be more deployments of large-scale IP
networks? Paul Doolan, Ennovate Networks Inc., Boxborough, Mass., says his company has not
seen great customer interest in the last year. "There is still a degree of
uncertainty in the standards world, and that always plays back to the customer community.
They don’t know what end is up, and they are always looking for a leader to do something.
From my point of view, the jury is still out."
Ruparel says Dialogic feels it must continue to support multiple protocols in the
interim. "We started with H.323 and a large number of customers today have H.323
deployments," he says. The company will support the older version of MGCP, because it
is used in the cable industry, as well as the new H.248. "We have also opened up some
of our low-level resources so our customers can plug in any call control they want: H.323,
SIP, MGCP or proprietary."
Seng-Poh Lee, director of product marketing, Convergent Net-works Inc., Tewksbury,
Mass., says his company has adopted a relatively proprietary solution for its softswitch
to maximize scalability. At the same time, it has kept its architecture open to protocols
such as MGCP and its successors. If certain scalability issues in MGCP have been resolved
in the new H.248, "that is the one that is going to win," Lee says.
Charlotte Wolter is infrastructure editor for PHONE+ magazine.
JAIN: Write Once, Deploy Everywhere
Palo Alto, Calif.-based Sun Microsystems Inc.’s Java in Advanced Intelligent Networks
(JAIN) is a Java-based protocol interface that is designed to allow telecom applications
to be portable to multiple platforms. As long as a vendor creates a JAIN interface for its
platform, an applications created with JAIN can run on it.
There are JAIN interfaces for various signaling system 7 (SS7) protocols, as well as
for Internet protocol (IP). There is also a JAIN interface for, operation administration
and management (OAM), a protocol that gives information about a network’s elements and
operational procedures, such as what to do in case of a timeout.
JAIN is designed to allow applications to interface, not just to different
implementations of SS7, but also from SS7 to IP networks. An example is a follow-me
application, which may first ring two phones on the public switched telephone network
(PSTN), then attempt to contact a user via e-mail. JAIN makes it much easier for the
application to operate across both kinds of networks.