IP Centrex (or hosted PBX) represents a major revenue opportunity for
service providers over the next several years. Industry analysts expect
the IP Centrex market to grow from approximately 13,000 lines in 2001 to
approximately 10 million lines by 2008, partly from migration of the
installed base of traditional Centrex users. IP Centrex is being deployed
by traditional service providers, competitive carriers and long-distance
service providers seeking to expand revenues by offering business
services.
Key drivers for the IP Centrex market are the generally recognized
benefits of Centrex services including lower upfront capital outlays and
ease of use and maintenance versus a customer-owned, PBX-based system.
Additionally, there are the benefits of IP telephony such as the lower
costs of sharing a single, packet-based, broadband access system that can
carry many simultaneous calls (and data packets) instead of dedicated
copper wire pairs for each endpoint; the ability to build groups of
Centrex users irrespective of their individual location who can benefit
from features such as abbreviated dialing; and the large reduction in
overhead to effect "adds, moves and changes."
An IP Centrex or hosted PBX system is a packet-based enterprise phone
system in which the IP-PBX or call manager function is owned and operated
by a service provider. IP phones are installed on enterprise user
desktops, either provided by the enterprise or by the service provider.
Typical IP Centrex systems employ SIP or H.323 to make calls and use
"internal" supplementary Class services as if they had their own
PBX. Currently, two basic types of IP Centrex solution are used:
- GR303 gateways, which provide an IP front end to a Class 5 switch,
thus making available the full range of (several hundred) Centrex
features; and
- Softswitches, which replace the CO's call switching function, and
hence make it possible for new and competitive "pure IP"
carriers to quickly introduce new services in competition with the
RBOCs.
In each case, packetized voice traffic travels over both the enterprise
LAN and the service provider's IP network. The enterprise LAN is owned and
operated by the enterprise and guarded by sophisticated firewalls that
protect the network from users on other networks. All packets traverse the
firewall, which examines each and rejects those not meeting specified
criteria. Consequently, the subscriber's LAN is made inaccessible to the
service provider, which makes service-affecting problems occurring on the
subscriber's LAN invisible.
Since the provider's and the subscribers' networks can be sources of
quality degradation, both networks must be visible to the service provider
to provide an "end-to-end" guaranteed or specified service
level, which for this article is the call quality level. Although a
service level agreement (SLA) may be in place to ensure the quality of the
service at the demarcation point between the two networks, this does not
guarantee the quality of the "inaccessible" enterprise network.
There is also an implied responsibility for the quality of the IP Centrex
voice that the service provider must meet, even though voice quality may
be affected by problems on the enterprise LAN.
USING CALL QUALITY MONITORING AGENTS
Given our scenario, how can the IP Centrex service provider see through
the enterprise firewall into the subscriber's network in order to maintain
the quality of the service? This dilemma is a significant concern to IP
Centrex providers and subscribers and has led to at least one major
service provider delaying their IP Centrex deployment and another risking
profitability by installing specialized hardware probes on the customer's
LAN.
A method gaining significant support is the use of call quality
monitoring agents that can be embedded directly into enterprise IP phones.
These are capable of sending extended RTCP reports (RTCP XR)
to provide call quality and diagnostic data in real time during the
call. Then, the service provider collects this extended information and
makes its own measurements at the demarcation point using agent
technology. In this way, any problems related to the packet stream will be
apparent and detectable in real time by the service provider, allowing
them to be proactive in reporting and resolving problems. Further,
comparison of call quality levels reported by the subscriber and measured
within the provider's network supports quick problem isolation, i.e.
"is the problem on my network or yours?"
RTCP is the "Real Time Control Protocol" implemented in
conjunction with RTP, the "Real-time Transport Protocol" (see
RFC 1889) used for voice, video and multimedia applications transmitted
over an IP network. RTCP reports are routinely sent every 5-10 seconds by
each IP phone during a call, traversing the same route as the RTP packets.
This approach solves the firewall problem nicely, as firewall routers are
already configured to allow both RTP and the associated RTCP packets to
pass through.
The primary function of RTCP is to provide feedback on the quality of
the session through the use of Sender and Receiver reports. However, the
existing RTCP standard (RFC1889) has major shortcomings -- it does provide
some performance metrics, however, these are grossly inadequate for
monitoring the performance of VoIP services.
In order to address those shortcomings, Telchemy, Brix Networks, AT&T, UCSB, ShieldIP, University of Piere and Marie
Curia, UT Dallas, Ericsson and
others in the Audio/Video Transport Working Group in the IETF have
developed a new approach that extends the capabilities of RTCP reports.
Dubbed "draft-ietf-avt-rtcp-report-extns-06.txt" and available
on the IETF
Web site, RTCP XR supplements the statistics already contained in the
report blocks with other and more detailed statistics that are tailored
for the management of VoIP services. A complete list of contributors to
the protocol is below.
Important RTCP XR metrics include:
- Packet Loss Density
- Packet Discard Density (due to jitter)
- Burst Length (mS), Burst Density
- Gap Length (mS), Gap Density
- Round Trip Delay and End System Delay
- R Factor
- MOS - Listening and Conversational Quality
- Packet Loss Concealment type
- Jitter buffer type
- Jitter buffer size (Min, Current, Max)
These statistics are critical in managing the IP Centrex service. The R
Factor and MOS scores give a clear and simple grading of call quality. The
service manager doesn't have to guess based upon isolated discrete
statistics such as packet loss or jitter. The MOS or R Factor models
aggregate and correlate the relevant information to provide an extremely
accurate score for conversational and listener quality.
Other metrics are crucial to the troubleshooting process. For example,
the most significant problem facing voice or video quality on an internal
network is jitter. The discard density is a perfect indicator of jitter's
impact while the jitter buffer metrics allow the potential
misconfiguration of the jitter buffer to be detected. Packet loss burst
metrics provide insight into another critical problem -- even if the
overall packet loss rate for the call is low, there are often high density
periods of loss that can be very annoying to the listener.
This powerful combination of software agents embedded into IP phones
and real-time call quality/diagnostic data being transmitted between IP
endpoints using RTCP XR provides a perfect solution for IP Centrex
management. Service providers have the ability to monitor the quality of
calls made by every IP phone, to measure quality at the desktop, to be
immediately aware of problems affecting quality and to have the lowest
possible cost for management infrastructure.
(Editor's note): A
complete list of contributors to the RTP Control Protocol Extended Reports
(RTCP XR) draft as taken from the draft:
Kevin Almeroth, UCSB
Ramon Caceres, ShieldIP
Alan Clark, Telchemy
Robert Cole, AT&T Labs
Nick Duffield, AT&T Labs-Research
Timur Friedman, Paris 6
Kaynam Hedayat, Brix Networks
Kamil Sarac, UT Dallas
Magnus Westerlund, Ericsson
Bob Massad is VP of marketing for Telchemy. Telchemy is a technology
leader in the market for Voice over IP call quality monitoring and Quality
of Service (QoS) management, and the "first" company to model
the effects of time-varying-impairments. For more information, please
visit their Web site at www.telchemy.com. |