Streaming - A Practical Guide to Video and Audio Compression - page 361

whose connection is running much slower. That content may then be cached there for a
while on the grounds that it is likely to be needed again soon by another user.
Following the terrorist attack on the World Trade Center in New York, all the news-
based Web sites found that their serving capacity was exceeded by orders of magnitude.
As a consequence, some of them now have an arrangement with Akamai to deliver their
sites through the edge servers at times of extremely high traffic. This has proven to be a
very good solution to providing Web-based coverage of other major events, such as the
Olympic games, other international sporting events, and the wars in Afghanistan and Iraq.
So the strategic opportunity that Akamai is exploiting is a service-based offering.
You can rent capacity from them when demand for your content exceeds your ability to
serve it.

20.12.2 Kasenna

Kasenna operates in a different way from Akamai. The company is interested in selling a
solution for you to deploy. You buy the hardware and software from Kasenna and run
your own installation. This is not the same as the service model that Akamai promotes.
Typical applications for Kasenna systems are major TV services to local community areas,
corporate organizations, campus TV systems, and distribution of pay-per-view services
within a condominium.
Kasenna also provides an interesting technical solution that pre-caches just the first
part of a large file on the edge servers. This is immediately delivered to the user as soon
as the user connects while the back end requests the remainder of the file from a central
repository.
This has the advantage of making all of the content seem to be available immediate
at all of the edge servers without the need to replicate the entire archive and install mas-
sive disk stores everywhere.

20.12.3 Which Edge Server Should You Choose?

If the content is not already fully cached in your nearest edge server, Kasenna's perform-
ance might be slightly better in terms of an “instantly on” start-up of a video-on-demand
stream when compared with Akamai. The Akamai solution is popular, reliable, and it
serves a large number of users simultaneously and efficiently from the same source by
virtue of the major back-end network support.
If you have a large number of users in several locations, you might even benefit
from a hybrid approach in which multiple Kasenna networks are set up with central
repositories but are connected to a central publishing system via the Akamai network.
This might seem expensive as an approach right now, but if the current trend for deliv-
ering TV services through the Internet continues to expand, some kind of major architec-
tural shift will need to happen because the TV viewing audience is far larger than the
Web-browsing audience. The bandwidth requirements are also magnified for delivery of
a predominantly streamed video experience rather than a few pages of text and graphics
to each viewer.