[*] VoIP SIG + SIP.edu call minutes - May 17, 2007

[Bill Reid]

I thought the Asterisk group might be interested in these notes.

-- Bill

_____________________________________________________________________

Notes for VoIP SIG+SIP.edu Conference Call - May 17, 2007

*Attendees*

Chris Casswell, MCNC
Todd Edwards, Wake Forest
Dave Laurentino, Research In Motion
Deke Kassabian, University of Pennsylvania
Candace Holman, Independent
Phil Kulick, Penn State
Jonathan Tyman, Internet2
Chris Trown, University of Oregon
Garret Yoshimi, University of Hawaii
... and several others ;-)

*Discussion*

Today's call features Bill Rich, the president and CEO of Pingtel,
providing a overview of open source SIP software. Bill starts by talking
about a recent survey that he completed for ACUTA on VoIP deployment,
which included a question about open source. The results indicate that
while a large number of respondents are considering open source
applications, 54% are not. A similar survey from three years ago
indicates that there are greater numbers of people interested in open
source now. Bill feels that this is representative of the early state of
VoIP deployment, where people are starting to experiment but have yet to
fully investigate open source options.
Bill talks about the Internet 2 reference architecture for VoIP, which
he feels is a strong intersection between open source communications and
VoIP. The reference drives what he calls the decomposed model, which
allows organizations to build their own solutions using best-of-breed
components in a vendor-neutral manner. This enables groups to deploy
services at their own pace and encourages innovation at every level of
the chain. Call control, networking, applications, and devices can all
be configured and combined in ways that were previously impossible and
in a manner that is diametrically opposed to the traditional TDM
hierarchy, which focused on tightly integrated single-vendor solutions.
This traditional approach, in Bill's opinion, offered limited
flexibility and innovation as well as elevated costs. The arrival of
open source communications systems has led to systems based on standard
protocols with more choices for handsets and applications, as well as
lower costs and quicker development cycles. Bill notes that some
traditional vendors are developing VoIP solutions, but they are still
based on the single-vendor models and often use proprietary or modified
protocols, closed software, and limited choices for hardware.

One key element of open source is that the software is freely available.
Customers can set up and configure the software on their own at no cost,
or purchase supported versions for lower cost than traditional software.
The open nature of the software leads to faster development and more
innovation. Bill notes that there are several myths regarding open
source, which he debunks. The open nature of the software leading to
security risks is one such myth; the idea is that since anyone can
contribute to the code it is easy to add malicious code. In reality this
is unlikely, as projects are centrally managed and contributions are
examined thoroughly before being integrated.

Another myth is that the software is of lower quality since it is free
and nobody makes money from developing it. In reality, most large open
source applications are funded by companies who pay developers and
perform QA work on the resulting software, often indemnifying customers
to ensure trust.

A third myth is that using open source software ends up being more
expensive due to development and support costs. Bill notes that most
major open source applications are available as commercial packages with
support, and users are only on their own if they choose to be.
Management tools for open source software have also gotten quite good,
with most offering GUI applications rather than the traditional
command-line interfaces.

Bill discusses the two major open source SIP projects: SIPfoundry and
Asterisk. SIPfoundry is an IP PBX application sold as a commercial
product by Pingtel and meant to serve as a unified SIP-based
communication core, using SIP terminals and extensions. The SIPxchange
software is sold by Pingtel on a subscription model, where customers pay
a certain amount per server per year, up to a particular number of
users. At the end of the subscription, the organization can continue to
use the software at no cost, but would lose support. Pingtel performs a
variety of testing and integration work with third party phones and
gateways, and employs most of the SIPfoundry developers. The product is
available as software only or as a complete package with all necessary
hardware, handsets, and servers.

The other major open source SIP product is Asterisk, which has been
around for close to seven years and is widely known and used. Asterisk
is popular with a number of small businesses, though there are plenty of
larger deployments as well, and is often used as a stand-alone voicemail
application. It is mainly commercialized by Digium, whose business model
focuses on selling cards to work with Asterisk. Other open-source
packages mentioned by Bill include SER, which is a SIP router, and
FreeSWITCH which is an outgrowth of Asterisk.

Bill offers a real-world example of costs for Pingtel's SIPxchange ECS
product. For 25 users, with hardware, a media gateway, the SIP server,
Polycom 430 phones, and a software subscription, including installation
and support, the cost would be $300 per seat. Moving up to 200 users
brings the cost down to $258 per seat, which is very cost effective
compared to traditional solutions. Bill also notes that the major cost
is for the IP phones, with the Polycom offered as a very
middle-of-the-road solution. The real question, he says, is if the
open-source SIP solution is ready for deployment in mission-critical
applications. Bill feels that this answer is a definitive yes.

All of Pingtel's software goes through traditional software development
practices, with releases in phases, QA, and faster release cycles of
three to four months. The majority of common features from other
communications software are available, as is support, a suite of
management tools, and the ability to integrate with technologies like
SOAP and LDAP. Current customers include a variety of colleges with
deployments from 500 to 2000 users, with plans to scale to 3000 total.
Amazon is also a customer, running their Seattle offices on Pingtel
systems. Bill says that there is probably a limit of about 10,000 users
on one server, but at a cost of $5,000 to $7,000 per server it is not
difficult to add more servers while remaining less expensive than
traditional systems.

Susan from Boston University asks about security measures in SIPxchange.
Bill says that both products incorporate TLS, and secure SIP signaling
and media available by the end of 2007.

Deke from Penn asks about the ability to do bridged line appearance and
busy indicator across multiple lines. Bill says that BLA is available in
SIPfoundry 4.0, which is scheduled for the end of summer and will work
across differently phone types. Certain phones will things slightly
differently, and initial support will probably be for Polycom and SNOM
phones.

Garret asks about installations for over 10,000 users. He's seen a lot
of deployments in the 100's and low 1,000's, but asks if anyone is
seriously looking at 30,000 to 40,000 line implementations. Bill
believes that this is the case, but would like to qualify his statement.
There are currently live implementations of greater than 5,000 lines,
and he's confident that 10,000 would work on high-availabiltiy
load-balanced servers. There is also work done on isolating specific
aspects, such as media or call center loads, onto different servers. In
lab tests these systems can handle 20 to 25 calls per second, so two or
three systems would work for 30,000 people. Voicemail and dialing plans
are another concern, but it's possible to have multiple systems using
one voicemail system.

Finally, Bill is asked about E911 support and lawful intercept. He says
that E911 is supported, but lawful intercept currently is not as in the
past it has been thought of as happening once traffic has moved to a
public network. He understands why this would be desirable, though, and
says that it is being worked on.

- - - - -end