AOL Developing Cheap Switch for Audio Streaming

legaleagll writes: "According to a Fortune magazine article and a follow-up article on ZDNet, AOL is developing a cheap switch that can handle streaming audio for 10,000 users, versus current technology of 100 - 1,000 users per box depending on expense of system. The code name for the product is Ultravox and was apparantly spurred into existence because RealNetworks is now offering internet service for cheaper than AOL. I'm a little skeptical because I'm not sure how the use of an intelligent router would eliminate the need for the expensive systems to stream the audio. Wouldn't moving the software for streaming onto the router make for a more expensive router and still require the expense a box outside of the router anyway?"

Comments

Effectiveness

[kwishot]

"I'm not sure how the use of an intelligent router would eliminate the need for the expensive systems to stream the audio. Wouldn't moving the software for streaming onto the router make for a more expensive router and still require the expense a box outside of the router anyway?"

I'm assuming that for these big operations that have multiple sites for streaming, it works something like this: You have a source for the stream, which no end-users ever touch. The job for this machine is to feed the routers, which can be in various areas of the world or whatever. That way each router has a dedicated "in" stream so that it can feed the masses. In a smaller operation where multiple sites aren't needed, this piece of hardware isn't needed either, so this thing actually does save money for an establishment that can take advantage of it's potential.

Why they want to do this

[Smallpond]

"I'm not sure how the use of an intelligent router would eliminate the need for the expensive systems to stream the audio."

Currently 1000 users == 1000 streams. An intelligent switch looks at the content, says broadcast one stream to the 1000 users. Sounds simple. Not, though. Check articles on "Layer 4" switches.

Multicast/unicast conversions

[silas_moeckel]

While niether article has anything technical I have worked in this industry for awhile now. I'm going to assume some things first off this is primarly for live streams only or simulated live as a router has no place mucking with vod streams it dosent have the memory to help besides some QOS things. OK now assuming it's live the best way to get the data out is multicast it but unfortunatly the Tier 1 ISP's cant figure out how to bill for Mcast and it dosent reach far enough like past an ISP and nearly never to anything like dialup or broadband it's a college toy. Now insert a router that can take multicast live streams and do a unicast conversion (it's not much harder than nat and just about everything that can load Cisco IOS can run NAT) Now that changes the math on the head end the servers only need to support the stream 1 time and the additional connections for authentication (non scalable multicast as real would term it) this would be a wonderfull thing to the source people and the end user (especialy if the cable head ends can convert back to multicast) and AOL has the size to get the teir 1's to play ball AOL is a huge installed base.

But like I said this is all conjecture on my part. It may just be them whalking the arrowpoint boxes again they are routers of sorts (load ballancers realy) that can accept drive space.