IP QoS? Yes. IP
Quality Of Service -- Present And Future
Go Right To:
QoS On The 'Net?
Implementing quality of service (QoS) in an IP network is a hot topic
these days. Today IP (Internet protocol) is ubiquitous -- the only
protocol to consider for new applications -- but it's a best-efforts
protocol. IP works great as long as there is enough bandwidth, but as soon
as there is any congestion, everything slows down equally. This limits
IP's ability to replace the telephone network, or even the majority of SNA
(Systems Network Architecture) data networks. Obviously delays have a
major impact on VoIP traffic. But, if you've just spent millions of
dollars on computer telephony gear to shave seconds off of call center
answering times, can you afford to add back seconds of delay while the
agent waits for a screen refresh? Giving priority to mission-critical
applications in the presence of network congestion is an equally important
motivation for a focus on QoS for IP.
In most cases, however, LAN bandwidth is fairly inexpensive, so there
is limited incentive to deploy QoS technology there -- especially if there
is appreciable first cost or maintenance cost associated with QoS. It's
usually easier to add more capacity -- higher speed segments or additional
Ethernet switches. A similar effect is occurring in the WAN. In recent
years, a vast amount of new fiber has been laid and long-distance
bandwidth, at least for service providers, is getting surprisingly
inexpensive. As a result, major data service providers offer frame relay
circuits with packet delivery guarantees and prices that are independent
of distance, at least within the United States. Other carriers are
offering IP services with latency and packet loss guarantees. But these
carriers are providing their guarantees using ATM equipment or plain IP
over-provisioning -- they are not deploying IP QoS, at least not yet.
So where is the big need for QoS technology? The biggest source of IP
congestion occurs at the edge of the enterprise, where an enterprise's 100
Mbps or Gigabit Ethernet LAN backbone connects to the WAN access link -- a
connection that may be, at best, a single 1.5 Mbps T1 pipe. It's the
access network that is the bottleneck today, and for the foreseeable
future. This is the first place where IP QoS technology will pay off, by
giving critical applications, such as voice, priority access to the
limited bandwidth pipe to ensure they have low delay and low packet loss.
EXISTING AND EMERGING SOLUTIONS
Several standards are complete or nearly complete, and diverse products
are beginning to roll out, creating a flurry of press releases and lots of
hype. But what IP QoS technologies are actually available? The first IETF
(Internet Engineering Task Force) effort
to be completed is a scheme called IntServ (Integrated Services). IntServ
introduced the Resource Reservation Protocol (RSVP) together with a
structure for making specific reservations of bandwidth at every router
along an IP path. Unfortunately, in order to reserve bandwidth, some
information must be stored in every router along the path and this must be
done for every session. There are also negotiations required at every
router along the path. If there are millions of users and billions of
paths, as in the public Internet, IntServ quickly becomes cumbersome. It
doesn't scale well.
To address this scalability issue, another IETF working group developed
DiffServ (Differential Services). This standard is nearly complete and
trial deployments are beginning. Under DiffServ, an individual user's
requests for bandwidth are aggregated into a few classes of service which
are provided from router to router on a per-hop basis. Traffic is metered
and shaped at the edge of a DiffServ network to conform to predetermined
service level agreements, so the core network is not overloaded. Each
traffic class is designated to receive a particular kind of treatment, and
statistically it is very likely to get this treatment. Because there is
only general information about a few classes of per-hop behavior stored at
any one point within an IP network, the DiffServ architecture for IP QoS
is expected to scale to any size. The downside is that there is only a
statistical guarantee of a specific behavior. The statistics can be made
arbitrarily good, but there's never an absolute guarantee.
In an entirely separate effort, the IETF has been working on
multi-protocol label switching (MPLS). MPLS provides virtual-circuit-like
connections through an IP network -- a major aid for traffic management
and therefore useful in support of QoS objectives. Basic IP is a
connectionless protocol. This means each packet gets to its destination by
whatever is the best route at that moment. In this, IP differs from
protocols such as X.25, frame relay, and ATM, which are
connection-oriented and establish a path or a virtual circuit between a
source and destination. In a connection-oriented network, all packets on
one virtual circuit use the identical set of links and switch/routers.
With a limited set of virtual circuits, provisioning problems are more
tractable. Packets can be sent along an explicit path different from the
one constructed by normal IP routing. Alternate virtual circuits can be
pre-provisioned to provide rapid cut-over in the event of a link failure.
And as long as bandwidth is provisioned for each virtual circuit, the
circuit can provide service guarantees. MPLS provides all of these
benefits for IP traffic, merely substituting the words label-switched path
for virtual circuit.
Today it is possible to map IP flows to ATM virtual circuits and
provide IP differentiated services in that fashion. MPLS can leverage
existing ATM switches, but can also work on any IP transport, potentially
reducing the complexity of maintaining both IP and ATM networks. Like
DiffServ, MPLS standards are very new and implementation experience is
limited. But these two new technologies are going to dramatically change
the communications landscape.
Within the next 12 months, we'll see the first IP differential services
offered. Initially these will be offered by mapping IP QoS to underlying
ATM services, but within the next 12 to 24 months, both MPLS and DiffServ
will be deployed and the cost of providing IP QoS will begin to fall. As
the technology progresses, the cost of providing special services for
applications such as voice and mission-critical systems will decrease
substantially. The statistics of utilizing DiffServ are such that
providing high priority or low latency for only 10 to 15 percent of a
network's traffic has a negligible impact on the rest of the best-effort
traffic, so in time, 10 or 15 percent high-priority packets may be
included with basic IP service. And since the total amount of traffic on
IP networks is growing at an exponential rate, voice traffic may well be
less than 15 percent of corporate traffic.
Today, a majority of corporations run traditional TDM (Time Division
Multiplexing) networks for voice traffic and use SNA networks for their
mission-critical data applications. But with the emergence of IP QoS
services over the next three years, any SNA-based application that
requires updating for any reason will be moved to IP as a simplification
The availability of IP QoS will directly affect our telephony
infrastructure as high-quality VoIP will be easily deployed. All PBX
vendors today have products or product plans that support IP connections
between different corporate sites. They also have IP-PBXs and Ethernet
telephones in the pipeline. While it's a nascent market, corporate
telephony is poised for major change in the next three years. And, as
IP-PBXs emerge, there will be additional opportunities for new
applications far beyond what PBX vendors are considering today. At the
very least, we will break away from the narrow-band (three KHz) telephony
we've lived with for the past 120 years. We may not need full CD-quality
for voice telephony, but VoIP will surely sound better than "toll
quality" does today.
The combination of high-quality audio and speech recognition will spawn
other interesting applications. For example, I'd like a business-oriented
chat service that could actually help me in my daily work. When I chose to
be "on-line," this virtual workgroup or virtual water cooler
application would provide a private, secure, IP-based, open-wire service
over which I could converse with co-workers, as if they were in the room
with me, even if they were in another country. I would want very
high-quality voice activity detection so that there wasn't any background
noise, only the voice of someone in my workgroup. And, I'd also like
high-grade speech recognition, so that I could speak a name and my voice
would be passed to only that person. All the needed technology exists
today, at least in the lab. With QoS in the corporate IP backbone, it
could be a reality in short order.
Internet telephony has come a long way from the hobbyists of 1995, but
the future is more exciting as IP QoS makes VoIP routine and unleashes a
host of new applications. Now is the time to figure out the new
communications services you want.
Brough Turner is senior vice president of technology at Natural
MicroSystems, a leading provider of hardware and software technologies for
developers of high-value telecommunications solutions. For more
information, visit the company's Web site at www.nmss.com.
dynamicsoft Intros eConvergence
dynamicsoft has introduced its complete eConvergence Server Solutions.
Led by the newly-announced SIP Application Server 1.0, the eConvergence
product family marks a ready-to-deploy solution suite based on SIP.
dynamicsoft's eConvergence product suite gives communication ASPs the
open, scalable tools to quickly develop and deploy enhanced services that
combine Internet telephony and other session-based services with the Web,
creating a robust experience for their customers. With this suite of
solutions, dynamicsoft addresses critical needs for deploying converged
networks, linking components such as call control, routing management,
subscriber management, and accounting/billing.
No. 530, www.itmag.com/freeinfo
GTE Internetworking Announces ESP Direct
GTE Internetworking announced ESP Direct, a new high-quality,
low-cost VoIP service that provides local inbound dial reach covering 80
percent of the U.S. metropolitan population. The service allows
"Enhanced Services Providers (ESP)" to deliver communication
services such as unified messaging, Internet call waiting, and IP fax to
their end users without having to make a large network infrastructure
No. 531, www.itmag.com/freeinfo
Trillium And Vsys Partner
Trillium Digital Systems and Vsys announced an integrated solution
to accelerate the deployment of Vsys' Necisa (Network Convergence
Integrated Switch Architecture) IP telephony platform. By integrating
Trillium's industry-leading H.323 software solution into the Necisa
Softswitch platform, Vsys enables network operators to quickly implement
carrier-class, end-to-end IP telephony solutions that meet the
performance, capacity, and scalability requirements of next-gen networks.
No. 532, www.itmag.com/freeinfo
ADC Intros Next-Gen Broadband Wireless Access System
ADC has announced the worldwide availability of the Axity
Broadband Wireless Access System, a carrier-class architecture for two-way
delivery of communications services over the MMDS wireless spectrum. The
introduction of the Axity system further enhances ADC's focus on the
access sector of broadband communications networks, and is an integral
part of the company's solutions for DSL, cable, and wireless.
No. 533, www.itmag.com/freeinfo
Phillips Group Predicts Triple Demand For Managed
According to a study from The Phillips Group, during the next
three years, more than half of U.S. companies will begin converting their
data, voice, and video traffic onto a single next-gen network utilizing IP
telephony. Eighty-one percent of the sites in these converged networks
will be connected using managed services from various service providers,
instead of private network, making a reversal of current trends in which
private lines are the dominant choice. Only 28 percent of company
locations currently utilize managed services for their data network.
No. 534, www.itmag.com/freeinfo
VocalData Releases IP Local Exchange Solution
VocalData announced the release of their complete IP local
exchange system, allowing service providers to deliver feature-packed IP
voice services and high-speed Internet access all over a single
packet-based connection to their customers. The architecture of
VocalData's solution functions in carrier or multi-tenant networks to
allow service providers and their customers to capitalize on the
efficiencies and cost-savings of a single communication network
infrastructure. The solution is also scalable, allowing service providers
to cost-effectively serve a limited number of customers, but with the
flexibility to grow into large network-wide configurations.
No. 535, www.itmag.com/freeinfo
IPeria Announces Service Node Version 1.2
IPeria has announced that version 1.2 of its Service Node and two
new communications applications, unified messaging and voice activated
voice e-mail, are generally available for deployments and trials by
service providers. The scalable, carrier-class IPeria Service Node enables
next-gen telcos, CLECs, wireless providers, and ISPs to deliver
value-added enhanced services to their customers. These services, in turn,
help providers produce revenue, increase average revenue per user, and
retain valued customers.
No. 536, www.itmag.com/freeinfo
Belle Systems Launches IMS 3.2
Belle Systems announced the launch of the latest version of its
Internet Management System (IMS). IMS 3.2 offers improved scalability,
faster data processing, and improved management of VoIP services. It is
designed for ISPs, telcos, and corporate organizations deploying VoIP and
other IP-based services. The latest version builds on the existing
capabilities of IMS, which enables the management of all aspects of IP
services from initial provisioning and deployment through to mediation,
rating, and billing.
No. 537, www.itmag.com/freeinfo
visitalk.com Offers PC-To-Phone Service
visitalk.com will extend its existing free global PC-to-PC audio
and video calling services by enabling its members to place calls from
their PCs to telephones worldwide. deltathree.com and MediaRing.com will
partner with visitalk.com to provide the PC-to-phone calls, which will be
initiated from the visitalk.com Web site. PC-to-phone calls will
complement visitalk.com's existing IP telephony services and offer members
new options for their long-distance calling.
No. 538, www.itmag.com/freeinfo
Net2Phone To Launch YAP
Net2Phone has introduced a line of hardware solutions designed to
enable a wide range of users to easily make Internet phone calls. The new
hardware line is called "YAP -- Your Alternative Phone," and
will be sold worldwide to consumers and enterprises through retail
superstores, resellers, and VARs this year. YAP Business solutions will
provide multi-line communications solutions suitable for small businesses
and larger PBX corporate environments. YAP Consumer products will allow
end users to connect any regular phone into Net2Phone's Internet calling
network through a standard phone line or an Internet-enabled PC.
No. 539, www.itmag.com/freeinfo
GRIC, HCI, USAN Deliver IP Services To Africa
GRIC Communications and HCI Technologies announced an agreement
with the United States African Network (USAN) to deploy multiple
value-added IP services for calling to and from selected countries in
Africa. The IP services will enhance and expand Africa's limited bandwidth
and improve the quality of intra-Africa calls, as well as overseas calls
initiated from select African nations. USAN, a prepaid international
long-distance provider, will be in a position to expand its service
offerings within the African continent. HCI is now growing its business
beyond the United States. GRIC will be extending its global reach into
Africa, providing multiple IP-based services to and from five continents.
No. 540, www.itmag.com/freeinfo
PassCall, Jacada To Deliver WAP Solutions
Jacada and PassCall Advanced Technologies will market and deliver
solutions that will provide immediate and complete connectivity to any
Internet and mainframe applications from any wireless device. Under the
agreement, PassCall and Jacada will work together to deliver WAP-based
solutions to AS/400 and mainframe customers worldwide. PassCall has
developed a technology that provides users of mobile phones and other
wireless devices with Internet access, utilizing existing Internet Web
sites. PassCall's GateWave technology enables access to any Internet site,
in any language, via any cellular device.
No. 541, www.itmag.com/freeinfo
eVoice Partners With Dialpad.com
eVoice has become the exclusive voice mail partner to Dialpad.com,
under an agreement that creates a one-stop shop for consumer voice
services online. Dialpad.com will offer eVoice's free home voice mail to
its subscribers, while Dialpad.com's one-click Web phone calling service
will be available to all users of eVoice's service.
No. 542, www.itmag.com/freeinfo
[return to the top]
QoS On The 'Net?
BY MARY EVSLIN
How do you manage not to have a "packet traffic jam" on a
road you don't own? You have to be a very smart traffic cop.
Let's take a step back for those who may be new to the Internet
telephony game, and discuss the necessities that make good quality. Three
things are needed, and if any one of them is missing, you have the makings
of poor call quality. First, the technology has to be top-notch and has to
have been tested in a serious lab. The gateway and gatekeeper software
have to be designed for more than point-to-point calling on a controlled
WAN (wide-area network) basis. The computers and operating systems need to
be industrial strength. Second, there has to be adequate bandwidth
available. And finally, someone has to manage the routing.
Global IP carriers who use the Internet for calls pose the greatest
challenge to gateway manufacturers. The Tier 1 global IP carrier can have
upwards of 100 points of presence (PoPs) and they are on a fully loaded,
truly distributed network that is susceptible to congestion.
The carrier must address all three causes of poor call quality. First
it must use equipment from manufacturers that have extensive experience in
the industry. It should purchase from vendors who are committed to
interoperability. It should not purchase equipment from any manufacturer
that is not committed to supplying onsite technical assistance. The
manufacturer gets as much out of this type of relationship as it gives. It
receives invaluable information addressing the challenges of those few
companies that are on the cutting edge of the new global convergence
Secondly the carrier needs to test any potential termination sites for
adequate Internet bandwidth. There are unfortunately some places in the
world where there is simply not yet enough bandwidth to assure quality
real-time voice and fax. The carrier building the network should test
extensively before deploying gateways to see what the latency is and how
much packet loss there is to a certain location. It may be that Internet
telephony cannot happen to a location until new satellite connections or
new fiber are available.
It is also up to the carrier to deal only with global Tier 1 ISPs that
have interconnection relationships with other Tier 1 ISPs. The Internet is
no longer owned by governments and universities. It is NOT a
not-for-profit organization. It is made up of many for-profit companies
that interconnect with each other. I find it hard to describe the
Internet; maybe the only thing that makes it the Internet is that I pay --
both at home and ITXC pays at the office -- not by the minutes, but for
access. I pay $19.95 at home and the company pays thousands of dollars a
month each to several Tier 1 ISPs for huge access pipes. The wonderful
thing about the Internet, besides the creativity it has spawned, is that
access gets one anywhere instantly.
For a global carrier, it is essential to do business on the Internet.
Not simply advertising, but actually providing transport on the Internet.
The cost of calling between G7 nations is decreasing at a very fast pace.
It is coming down much more slowly between developing countries, but it is
still coming down. The Federal Communications Commission and the World
Trade Organization are applying pressure to governments to encourage
competition and to lower settlement rates. In order to keep margins up,
carriers need to manage their costs. The cost of transport is a large part
of the carrier's cost structure. Using the Internet dramatically changes
costs. Likewise, it dramatically changes quality issues.
The final condition for good quality calling is the management of the
routing to eliminate or reduce latency and packet loss.
The Internet is truly a distributed network with no single points of
failure. A business following Internet economics also has no single points
of failure. In the early adopter Internet telephony industry, let's be
honest, there is no such thing as a gateway or router that cannot fail.
There is no such thing as a part of the Internet that is not subject to
occasional congestion. So the answer lies in redundancy.
The carrier must have multiple gateway operators at each terminating
site. Each gateway installation should ideally have its own ISP and
central office. The carrier itself should have more than one Tier 1 ISP.
The carrier should have redundancy in its own hubs and redundancy in its
routers and switches. It should have the ability to provide backup managed
IP or PSTN if needed to cover times of extensive Internet congestion. No
single element of a global Internet telephony carrier's network is up to
the 999s of reliability, but the total of all the elements (put together)
In order to manage this redundant network, the carrier absolutely needs
a 24x7 network operations center (NOC). The NOC should be using tools to
continually test for quality. If the NOC detects that quality (call
completion, call duration, Internet congestion, or latency) is not up to
strict standards, it needs to have the ability to dynamically reroute
calls to improve quality. Unfortunately, there are no tools that can be
purchased off the shelf. Today's leading Internet telephony carriers have
to write their own tools. In order to write an intelligent tool,
intelligence must first be gathered. Tier 1 Internet telephony carriers
have developed databases of call durations and completion rates and what
the results of ongoing "ping" tests should be to certain
locations at certain times of the day. These tools notify the carrier when
they are not achieving a prescribed standard of service.
Staff members who see degradation in the network do several things.
They notify the terminating company that there is a problem and begin the
troubleshooting process. At the same time, they reroute calls to other
terminators or through other transport mechanisms until they solve the
problem. It is not advisable for carriers to leave this observation and
intricate rerouting to affiliates as some clearinghouses do. It takes
highly trained and experienced staff to handle this type of issue.
Maintaining QoS today is a skill that is being developed by
cutting-edge Internet telephony carriers. It is accomplished by a
combination of procedures, policies, sophisticated routing tables, and
constant vigilance by trained staff and newly developed software tools.
The industry will eventually come to agreement on a set of standards for
QoS, but today it is the early entrants into the market who are creating
Mary Evslin is vice president of marketing and customer services at
ITXC Corporation. ITXC is the service providers' service provider and the
operator of infrastructure that voice enables the Internet. The company's
WWeXchange Service provides phone-to-phone wholesale call completion for
carriers and resellers and has been chosen by 10 of the top 12 U.S.
facilities-based carriers, leading European competitive carriers, and PTTs
worldwide to complete their customers' calls. For more information, visit
the company's Web site at www.itxc.com.
[return to the top]