Thursday, April 3, 2008

Network Savvy

One things I've been meaning to point out for a while in this blog...and an email I got today reminded me that I need to... was to talk about a key criteria for picking out a High Definition MCU. And that is, how smart is your MCU when it comes to bandwidth utilization. Lets look at two different ways to do it.

One way to do it is to have every endpoint send a 720P image to the MCU. Then, because a lot of people want a CP layout, the MCU has to take that 720p image and reduce it in size to be able to stitch it into new picture that is also 720P to send back out to everyone. So, for example, four endpoints send the MCU 921,600 pixels of information. Say, we've chosen a Quad split. (probably the most popular layout) We have to take 4 x 921,600 pixels and compress it down to just one 921,600 pixel picture to send back to everyone.

Another way would be for the MCU to request each endpoint to send less than 921,600 pixels. The next most popular resolution would be 4SIF or 704x480 or 337,920. pixels. With a smaller image being sent, the MCU doesn't have to spend compute time compressing the picture down to fit into one composite 720p image.

Both ways are applicable, but only one of them is network efficient. In the first example four endpoints each send 921,600 pixels...or a total of 3,686,400 pixels. We only need 921,600 to send out, so, tremdousely oversimplifying things, the mcu has to throw out 2,764,800 pixels... or essentially 75% of the information sent to it. In the second example, four endpoints send 337,920 pixels each for a total of 1,351,680...only oversending 430,080 pixels.

Even though bandwidth has gotten cheaper, I know that all network managers would certainly pick option 2, because it would reduce up to 75% of my inbound traffic to the MCU. Outbound would stay the same in both scenarios.

So, which MCU is smart enough to reduce network utilization? and which MCU doesn't needlessly have to shrink down 720p images down to 4SIF images? Only one, Polycom. Our main competitor Codian, takes the Mack truck approach of wasting lots of resources... while Polycom takes the Toyota Prius approach of using what resources are needed. I think most people would pick the more elegant, more network friendly approach.

1 comment:

Sean Lessman said...

the Codian MCU actually does both methods, when it has "Video receive bit
rate optimization" enabled it will actually flow control down and
endpoint when no one is viewing them in a large square.

Sean

 
ss_blog_claim=696d2076fda43d78281e6dc80d7c177d