Content Distribution including P6
Content Distribution Is Next Big Thing
By Max Smetannikov, Inter@ctive Week
September 4, 2000
Some carriers are shelving plans to deploy multicast
technology as a way to stream video and voice over the
Internet as they turn their attention to content
distribution, a business model that offers instant
gratification to service providers seeking to launch
streaming in the near future.
Heavy users of multicast are leery of this trend, and
some major players are hopeful that recent
developments will make the technology easier to use.
But service providers actually making choices on how to
stream content are voting with their wallets.
Increasingly, carriers such as PSINet (www.psi.net) and
Web hosters like Exodus Communications
(www.exodus.com) are seeking to strike deals with
content distributors instead of building overlay networks
that support multicast. PSINet took an equity stake in
Streampipe. com, formerly Netcast, and Exodus
invested in Mirror Image Internet (www.mirrorimage.net),
both of which are content distribution companies.
Multicast, formally defined as a way to communicate
between a single sender and multiple receivers on a
network, is a protocol for streaming content over the
Internet.
A typical arrangement involves routers that can read
multicast-enabled packets and thus reproduce audio or
video messages to single or multiple computers. But
most routers on the Internet are not enabled for the
technology. Multicast-enabled networks today are like
atolls in an ocean of networks choosing not to deploy
the technology.
From the 'edge'
Streaming voice and video via a content distribution
model is different. Streaming servers transmit voice and
video to users from the "edge" of the network. Unlike
with multicast, content distributors can deliver streams
to users on hundreds of networks with comparable if not
superior quality. The means of delivering content to
these streaming servers varies from satellite distribution
to proprietary routing algorithms designed to ensure
delivery and quality.
"With distributed servers scattered around the world, you
get kind of the same result as multicast," says Ed
Frankenberg, director of Web products at PSINet. "I
would say that multicast does have a role that is
associated with one-time events that are popular to
watch, such as the Super Bowl."
PSINet doesn't have a multicasting product at this time,
though it has experimented with the technology for a
while. PSINet offers boutique streaming services based
on unicast - streaming multimedia through a server -
which it complements with Stream pipe's servers. In
essence, the company is scaling its unicast offering
beyond its network by relying on a content distribution
model.
Because multicast content reaches people only on a
network where the technology has been enabled,
interconnecting with the likes of PSINet would give a
content provider access only to eyeballs connected to
PSINet backbone. Content distributors, on the other
hand, promise content providers literally the world, by
deploying streaming servers in a multitude of networks
and coming up with innovative methods of streaming
their traffic to these "edge" servers, while avoiding delays
and packet loss associated with multiple hops on the
network.
PSINet already offers a cobranded streaming product
with Stream pipe, and plans to dramatically scale its
joint content distribution network in the next six months.
PSINet's multicast plans have been shelved indefinitely.
The rationale for this new trend seems to be evident.
Large carriers like AT&T, Genuity, Sprint and UUnet,
which have either already developed multicast as a
product or are pursuing such plans, have failed to
establish an alliance that would link users on major
backbones into one multicast cloud.
The economics of backbone relationships, which
manifest themselves in peering contracts, have removed
any economic incentives for backbone providers to
deploy and interconnect multicast networks. Content
distributors such as Akamai Technologies, however,
solved the peering riddle by creating
network-independent content distribution footprints, the
model that carriers like PSINet seem to be embracing to
offer streaming services in the immediate future.
Skeptics say this might be the first sign of multicast
being shoved into the dumping ground of great
technologies that never found a market, like wireless
cable.
"It seems like every year is to be the year of the
multicast, and that continues not to happen," says Alex
Benik, an analyst at The Yankee Group.
But just wait until these content distribution networks
have to perform under a real streaming load, warn
executives at Yahoo!Broadcast, Yahoo!'s broadcasting
service, which has been handling large streaming
operations for a number of years.
Yahoo!Broadcast's operation had to purchase dark fiber
to connect its Dallas-based broadcast center with
several key backbones so it wouldn't be dependent on a
telecommunications carrier if it suddenly had to add a
big pipe for a large broadcast. The broadcast portion of
its network is simple. Yahoo!Broadcast owns the last
mile, connecting its broadcast center with key multicast
partners such as UUnet.
Yahoo!Broadcast's executives view content distribution
strategies as an interesting alternative to multicast, but
not a real competing technology. Yahoo! Broadcast has
been using multicast for five years, and company
executives think it is the way to go.
"The economics of proposals that I got from content
distribution companies just don't hold up to what we have
got," says Brokaw Price, director for business
development for multicast and broadband initiatives at
Yahoo!Broadcast.
Price says his company has a relationship with 70
carriers that have multi cast-enabled their networks. He
believes the content distribution model won't scale for an
operation as big as his. "Imagine managing 15,000
servers - all different platforms with all different software
- in order to do one broadcast," Price says.
Indeed, until recently, in order to do a single broadcast,
content distributor Digital Island had to program each
individual server to receive a certain broadcast, a manual
process that took days to execute. However, companies
such as FastForward Networks are eliminating that
problem using software. FastForward
(www.fastforwardnetworks.com), which sells software to
manage streaming servers deployed in different
networks, has made deployments both with PSINet's
Streampipe and Digital Island.
"Using our technology, we can help customers like
Streampipe make content distribution scalable for real,"
says Abhay Parekh, FastForward's founder and chief
executive.
But even if the content distribution model can in fact
scale to Price's satisfaction, there is still the question of
economics. Players like Streampipe and Digital Island
still have to buy bandwidth from the same people Yahoo!
Broadcast buys it from - large backbones like UUnet.
Because of the sheer size of its operation,
Yahoo!Broadcast can get massive discounts that
wouldn't be available to the likes of Streampipe.
This economy makes multicast a technology of choice
for large content providers. Instead of paying content
distributors a premium for streaming out their content
over large backbones, large players can cut deals with
backbones directly and eliminate the middleman.
Unless, of course, content distributors deliver on the
promise of better quality of streamed content, which
would make their networks more desirable than
multicast.
If UUnet (www.uu.net) believes that content distribution
can offer higher quality for streaming media services than
UUcast, the company's multicast service, multicast's
future as a premium solution for large streaming
operations may be shaky. If the economics of the two
services become comparable, content providers may
start to look at multicast as a low-quality alternative to
content distribution.
zdnet.com |
