Manifest Technology Blog
-- Site:
| Articles
| Galleries
| Resources
| DVI Tech
| About
| Site Map
|
Articles:
| PC Video
| Web Media
| DVD & CD
| Portable Media
| Digital Imaging
| Wireless Media
| Home Media
| Tech & Society
|
PC Video: |
PC Video Articles |
Video Software Gallery |
Video Editing Resources |
PC-Based Video Servers: Princeton Server Group
(9/2004)
by Douglas Dixon
DIVA
PSG Beginnings
NJN / HD Servers
Team
Brainstorming
Markets
TV30 / Broadcast Servers
Classroom VOD
Growth
References
What's an entrepreneurial technologist to do when the $500 million start-up
that he is working at crashes and burns? For Jesse Lerman and Jim Fredrickson,
the wild ride -- from the original spin-out of Sarnoff Real-Time Corporation to
DIVA Systems in the fall of 1994, to DIVA's ultimate collapse in fall 2002 --
suggested a better model, not only for developing core video server technology,
but also for building a company. Especially after the dot-com bust, the approach
they have taken with their new company, Princeton Server Group (www.princetonservergroup.com),
has been to eschew venture funding and big-bang growth. Instead, they are
self-funding, growing slowly, and taking advantage of Princeton area connections
to find staffing, support, and also customers for their company.
Their model has yet prove itself, but the contracts are beginning to roll in,
expanding the company's reach from local to regional contracts (including a big
one with New Jersey Network) to national accounts.
DIVA Systems Corp. was formed in 1995 to exploit the Princeton Engine
video processing computer developed at Sarnoff Corporation, in Princeton,
N.J. (www.sarnoff.com), and then
commercialized through its Sarnoff Real-Time Corporation (SRTC) spin-off.
Lerman was one of the first hires into SRTC, after graduating from Princeton in
1994 with a bachelor's degree in electrical engineering, and then served as
Director of Software at DIVA. Fredrickson, who received a bachelor's degree in
computer engineering from Ohio State in 1984, was a Sarnoff employee already
working on the video engine, and became principal engineer at DIVA.
DIVA's plan was to provide video on demand (VOD) services to cable TV
companies by leveraging its proprietary video server technology. DIVA was first
to market with a complete VOD service, and as of fall 2000 had systems deployed
in six cable systems in the U.S. with multiple system operators, including
AT&T Broadband.
However, DIVA had an expensive custom system, and the cable companies were
not enamored with DIVA's business model, licensing servers and content and
trying to share in the VOD revenue stream, so that through 1999 and 2000 DIVA
had needed to significantly revamp its business model to simply selling
equipment.
According to the Los Angeles Business Journal, DIVA incurred net losses of
$500 million over a six-year period while collecting just $39 million in
revenue, and still held only 15 percent of the VOD market. DIVA filed for
bankruptcy in June 2002. The end came in October 2002, after Gemstar-TV Guide
had originally agreed to buy the company for $40 million, and then backed out of
the deal days before the closing, after effectively shutting down DIVA's
business.
With the death of DIVA at the end of 2002, Lerman and Fredrickson also saw
the end of expensive, proprietary video processing systems, due to the
tremendous performance improvements of commodity personal computers. They
therefore set out to build a much less expensive solution for storing and
delivering video streams.
"The idea was that our system would be much more cost-effective,"
says Lerman, "higher performing, rock-solid, reliable, better than all of
the hardware-based platforms that are out there, and that are also very
expensive -- companies as big as 300 to 400 people trying to do something that
we can do with a very small company."
"So we took a grass-roots approach," he says, "which was to
start to build the technology first." By the spring of 2003 they had
succeeded in developing the base technology -- not just a demo or prototype, but
the core implementation of a video server using off-the-shelf PC hardware.
However, "we learned pretty quickly the reality was the cable companies
aren't really going to buy from a small start-up," says Lerman.
"There's a very high start-up cost for entering that space."
Instead of looking for venture funding, or attempting to staff up and build a
huge company overnight. Lerman and Fredrickson chose to take a step back and
re-consider the market for their technology. "We did not believe there was
any investment money available in the spring of 2003," says Lerman.
"We were not going to burn time trying to raise money." Instead,
"we can build some products and get customers, and really let that speak
for itself."
In the summer of 2004, PSG delivered a HDT2000 High-Definition broadcast
server system to New Jersey Network to provide similar flexibility in
broadcasting NJN's public broadcasting high definition content (www.njn.net).
"We have specific needs when it comes to broadcasting digital video
content," says Rick Williams, assistant director of engineering, "and
the folks at PSG have really filled a niche. Early last spring we were
challenged with coming up with a way to record HD content from our PBS network,
and they were able to come up with a cost-effective solution. As a state agency
we have to go through a stringent vendor purchasing process, and they were able
to meet it."
The PSG system captures, stores, and then plays back the HD content delivered
from the national PBS network. "It's very hard to do with HD content,"
says Williams, "because you have to have the HD infrastructure in place.
PSG was able to come up with a system that recorded content and played it back
out automatically. That really helps us; it's a labor saving device. The PBS
material now can fit within the schedule we determine. It also gives us a venue
for our own local HD content, which we can put into the PSG system."
But why PSG? "There are a variety of manufacturers doing this,"
says Williams, "but in this instance we needed somebody to help us bridge
the gap between where we are currently in our infrastructure build-out, and
where we are going to be in the next couple of years. Ultimately, we will
probably go with a much larger system, but then again as PSG moves along they
may be in that area at that point; but right now they really came in at the
right time and the right place." And the right price: "You can buy
these kinds of systems for hundreds of thousands of dollars," says
Williams, "and they came in at tens of thousands."
The PSG system is "working well," says Williams, "and actually
is part of a continuing process of testing and enhancement. As we require
changes they are able to keep up with those requests." And the exchange is
two-way: "They also provide us with new techniques. We're doing a lot of
beta testing for them."
Princeton Server Group was formalized in July 2003. The original founders
were Jesse Lerman as president & CEO, Jim Fredrickson as chief technical
officer, and Harini Subrahmanyam as principal engineer.
As they recognized the need to re-focus their target markets, Lerman and
Fredrickson also drew on their Princeton area connections to add marketing and
business expertise. "The original founders were all very technology
focused," says Lerman, "so we started to reach out and start
networking locally to grow the business side."
Lerman had kept in touch with John Romanowich, a former colleague at SRTC who
has been active in the Princeton entrepreneurial community, and who was in the
process of starting his own third company, Automated Threat Detection.
"John had helped with due diligence on Princeton Power Systems for its
investors," says Lerman (referring to another local start-up founded by
Princeton University graduates). "He thought with my ties to Princeton
University, our start-up would be a good candidate for Ed Zschau's technology
incubator." Zschau, visiting professor at Princeton from Harvard Business
School, developed the New Enterprise Lab incubator on Princeton University's
Forrestal Campus. PSG then began sharing space with Princeton Power in one of
the Forrestal lab buildings.
In addition, Romanowich introduced PSG to Steven Georges, who eventually
joined PSG as chief financial officer. Georges had worked in the media industry
since graduating from Harvard Business School with an MBA in 1984, having earned
his bachelors from the Wharton School at Penn in 1980. Georges began his career
at CBS, and then joined Time Warner in 1991 to co-found and then manage the NY 1
News regional news channel. From 1994 to 1997 he was vice president of business
development at Time Inc. New Media, and director of finance and administration
for Time Warner's News on Demand.
By 2003 Georges had started an independent consulting firm to provide
strategic financial leadership for media, technology and nonprofit projects.
"I was very tired of commuting to New York where my network was," says
Georges, "and I just put a nickel on a map around Princeton and said this
is the radius I wanted to look in. I had been targeting finding a team that felt
right locally. I liked Jesse: his values, his vision and creativity."
Romanowich also had previously introduced Georges to Paul Andrews, who became
PSG's vice president of product marketing and systems engineering. Andrews
graduated from the College of New Jersey with a bachelors in engineering and
then joined RCA (GE) Americom in 1984. He did engineering, operations, program
management, and product development for a succession of satellite companies:
joining Edgix, a satellite caching start-up, in 1996, co-founding SkyEx Networks
for satellite file delivery in 2001, and consulting to Loral Skynet on satellite
two-way broadband until December 2003.
This group then spent the summer of 2003 getting to know each other and
brainstorming on possible markets and business directions for the summer.
"It's the longest interview I'd ever been in," says Andrews.
"Starting in July we had a Monday morning coffee club. We got together
every Monday in person and talked about the potential for markets, and then it
snowballed from there. We also corresponded by e-mail just about every
day."
"We were looking for which markets felt right," says Lerman,
"where we think we could make an impact quickly, and generate early revenue
so we could self-fund the company."
"We went through a whole process where we were brainstorming all sorts
of ideas," he says, "some of which were really amazing ideas and could
have made a ton of money, but were really hard for a very small company to go
after. We still have those ideas, and they are waiting for the right time. We
are trying to grow at the right pace."
"We started realizing there were some very interesting markets that the
big players wouldn't go after," says Lerman, "because they were not
quite big enough."
"From my experience in the satellite industry," says Andrews,
"and the kinds of customers that came to us -- schools and big television
stations and small stations -- I had an archive of problems that the bigger
companies could not solve because there just was not enough market there, or the
technology was not right."
"What we found is that there are a number of markets that our technology
was not only a great fit for," says Lerman, "but this whole concept of
build once and sell to many different markets, which was something we were
looking at since we did have a very flexible technology."
The result of the brainstorming in summer 2003 was a clear strategy for PSG
to focus on developing on-demand services for the kinds of customers that
previously could not afford high-end proprietary solutions, such as
corporations, universities, nonprofits and communities.
"PSG has 'big time' experience with digital media technology and
business," says Lerman, "including cable VOD, satellite, NY1 News,
Sarnoff, and DIVA. Our technology approach (a pure software media server engine
that leverages standard PC platforms and IT technology) allows us to drastically
lower cost, go to market quickly with new features, and simplify digital media
systems. This has allowed us to solve problems for niche markets that could not
previously afford or easily leverage digital media technology."
"We started realizing there are a lot of opportunities that are
community-based," says Lerman. PSG is currently working with several local
organizations as its first customers. It is deploying classroom video on demand
systems for education (including the Princeton and Hopewell high schools),
program storage and scheduling servers for community broadcast stations
(including Princeton's TV30 cable station), video distribution edge servers for
education and emergency management (including NJN's Digital Classroom), as well
as high-definition broadcast servers for PBS affiliates (including NJN).
TV30, Princeton Community Television, went on air with the PSG B1000
broadcast server in late July 2004 (www.princetontv.org).
This kind of public-access broadcast station is known as PEG, for Public access,
Educational, and Government.
By storing programs on a server as digital video files, the PSG server allows
PEG stations to conveniently manage and schedule their entire broadcast day,
instead of having to juggle analog video tapes and switch sources among a group
of analog video recorders. "One of the things we were looking for was to
get away from proprietary boxes and tape decks," says Serge Suarez,
technical director at TV30, "and just go entirely over to basing our
program schedule on a generic PC and using MPEG-2 [digital video] files as the
content vehicle
One key advantage for Suarez is the standard PC platform: "Since the
product is based on a generic architecture, you don't have to worry about
shipping a box back to the factory to get a part fixed. If you need a bigger
disk or a higher network bandwidth card, you can just open the box and slap it
in there and you are done."
Even better, the PSG server uses a standard Web interface, so the system can
be controlled remotely. "It allows us to manage it much more
efficiently," says Suarez. "We can do all of our program setups from
home or wherever you happen to be at the moment."
Remote management is a particular blessing for station managers in bad
weather or emergency situations. "One person we spoke to would guess
whether school was cancelled or not before going home," says Georges,
"and then he would get up at 3 a.m, and if he was right he'd go back to
sleep, but if he was wrong he would have to drive in and change it."
The PSG broadcast server system for PEG stations is priced at $4950. It
includes 500 gigabytes of storage, enough to hold approximately 250 hours of
DVD-quality digital video. It stores and manages the program material, and then
plays it out to the cable system according to the preset schedule. The server
simply appears as a shared storage resource on the network, so it is easy to
access: you can just drag and drop files to it. "This is network-attached
broadcasting," says Andrews, "it looks like a storage server, and you
can use it for archiving."
PSG S2000 Digital Media Server
The PSG approach also allows a growth path for TV30. "It's open ended
because of the way they designed the software and the hardware platform,"
says Suarez. "Down the road we are going to be adding a video capture card
that will allow us to capture analog inputs under schedule control," such
as the current live monthly broadcast from the Princeton Arts Council.
PSG also offers a separate storage server so that stations can archive and
access their libraries of material. "PEG stations have a huge problem with
storage because VHS tapes are big," says Andrews. "They're starting
now to archive to DVD because discs are easier to store, but they still have the
problem of accessing it." The PSG S1000 and S2000 servers provide terabytes
of storage and up to 2 gigabits per second of throughput for simultaneous video
streams for prices in the range of $5,000 to $10,000 to $20,000, depending on
the storage capacity. These systems also offer dual power supplies and RAID disk
drives to provide protected storage.
"We're basically running our station with it," says Suarez.
"We have a handful of volunteers, and a very limited budget, so we have to
get something that basically runs by itself and doesn't require a lot of
tending."
"We're quite happy with it," he says. "It's exactly what we
wanted."
Broadcast stations are a good market for PSG. "We look for markets that
have a lot of need," says Andrews. "New Jersey is nice because it is
densely populated state with a lot of cable systems, and PEG channels (ninety in
the state). Some are run by schools, which intersects with another market:
classroom video on demand."
In the educational market, PSG can offer the same kind of centralized
archiving and administration of digital media on servers, but deliver the video
over the local area network to set-top decoder boxes in the classrooms.
"Schools have invested in digital networks," says Georges, "and
they want rich media, because kids learn visually. But when they bring media
into the classroom they're wheeling in a VCR." This is a nice match to
PSG's original focus in video on demand: "These are customers that could
not afford a classical VOD system," says Andrews, "and don't need all
of the other things you need to run a VOD system like billing."
The schools have already entered the digital media age, with network
connections and media departments. "Schools create a lot of content
internally," says Andrews, "all the baseball and soccer games, and it
typically gets lost. It goes on a videotape in a locker somewhere; it's lost
institutional memory. It's valuable stuff, since the way people learn is
building on the efforts of previous generations. If you are a kid actor trying
to learn your part in a Man of La Mancha, wouldn't it be great if you could go
back and see how it was done in previous years."
The system also needs to be easy to use. "We provide set-top boxes in
every classroom," says Lerman, "and the centralized storage server in
the data center. The teachers can point at the set-top box with the remote
control, just like Comcast, except all the content on the server would be
material that the school district has deemed important." Teachers also can
pull out selections from the stored material as pre-stored playlists to then be
used to augment the classroom curriculum.
Storing the content on-site not only provides more convenient access for
locally-generated material, but it also avoids the bottleneck of the Internet
connection. "People are trying to put a library of content out on the
Internet," says Lerman. "This will be completely unusable -- you can't
have 25 or 100 classrooms all trying to access the Internet at once. We're
bringing the cache inside the school so you now have high-quality instant
access."
Since the PSG system is compatible with off-the shelf set-top boxes, the cost
to equip each classroom rides consumer pricing, now down to around $100 to $200
per unit. Similarly, by using standard disk drives, the PSG systems can grow in
capacity with new technology. "Schools have an archive of existing video
tapes that they are able to ingest into our server," says Andrews.
"Our archive servers are multiple terabytes with twelve or more hard
drives. Early next year you could have over 10 terabytes."
PSG is deploying classroom VOD systems this summer for use in the new school
year, for the Princeton Regional School District, the Hopewell Regional School
District, and for Arizona State University-West media studies department. PSG is
also developing a digital multimedia system for the Numina Art Gallery at
Princeton High School for installation this fall.
"Princeton Server Group has taken an 'old fashioned' approach to growing
a business," says Lerman, "starting in the community, building
practical solutions, and growing through word of mouth. For example, Peter
Thomson from Princeton Regional Schools is an early adopter of PSG's Classroom
VOD suite. He recommended we speak to TV30, Princeton's community PEG-TV
station. TV30 recommended us to stations in Saddle Brook and Woodbridge.
Woodbridge introduced us to Edison and Metuchen."
And the growth continues: In August 2004, PSG delivered broadcast and archive
servers for the community station in Saddle Brook, N.J., and for the PEG station
in the town of Keizer, Oregon.
"We're self-financed, and cash-flow positive," says Georges,
"a little bit of money is going a long way. We really believe in it. We
like the idea of bootstrapping. It's like just in time; it lets you see where
the problems are quickly."
"It's easier growing this ourselves," says Lerman, "going
after the markets that may not offer the big fast reward. The technology itself
is suitable for the other bigger environments, and as we grow we'll pursue those
other markets." For now, however, "we're really happy about these
markets; It feels good, and we are contributing a lot to the community. It's
exciting to see the excitement of a PEG broadcaster who is really getting the
first taste of a digital playout -- for the first time seeing how easy this
stuff is -- and they're just so thrilled."
"PSG just happens to be in our backyard," says Serge Suarez at
TV30, "and they obviously have a lot of experience given where they came
from. Their philosophy and their offering is basically just what we are looking
for. When you call them, the entire office has very qualified and very savvy
people."
"It's nice knowing that there's a New Jersey company, a local company,
that can help us out," adds Rick Williams at NJN, "and they've really
been able to provide that niche. They're a small group of guys who have a very
productive output."
Princeton Server Group
www.princetonservergroup.com
Sarnoff Corporation
www.sarnoff.com
New Jersey Network
www.njn.net
TV30, Princeton Community Television
www.princetontv.org
|