CDN Scaling Strategies

Learn more about CDN quality at Streaming Media's next event.

Watch the complete presentation from Streaming Media West, B104. Fireside Chat: The State of the CDN Market, in the Streaming Media Conference Video Portal.

Read the complete transcript of this clip:

Neil Glazebrook: The winner in this market, certainly from the cost perspective, is the organization that can figure out how to fill in, you know, the valleys. Because building for the peaks is a very, very costly proposition. You know, the traffic levels in Chicago and London, Paris and Frankfurt, great, but, you know, if you build out to maximize those demands, what do you do in periods where you've got massive, massive amounts of capacity that's largely unused.

So, as much as it's a capacity challenge, I think it's probably more a traffic engineering challenge. And, for those software updates that don't have to be served, you know, by the nearest POP and where latency is not quite such a critical KPI, it's maybe serving that from a, you know, a location that has more capacity. So, it's really being, sort of having that level of intelligence that I think is gonna really help all of their organizations scale.

Dom:George, at Tulix, what about your capacity planning?

George Bokuchava: We have a very interesting approach, maybe a little bit different from a lot of other cities. And there were a few interesting articles a couple years ago in Streaming Media magazine about our approach, which actually, it's not the only approach, we have a hybrid approach. But, specifically, what we're doing, a lot of our traffic comes out from our main POP in Atlanta. We're in the process of building a terabit POP in Atlanta right now, with equipment. Equipment includes multiple stages, including, obviously, network stage, network equipment. Which is very important to have. It takes sometimes very long time to...

So right now we're talking about, 50-terabit capable of equipment doesn't mean it'll start obviously with the traffic right away. For more than ten years, on the other hand, at Tulix we neutralized our routing, optimized network. This is how we go to the live streaming services very quickly, because again, for more than ten years, we've been using a realtime routing optimizer, which we're optimizing all the live streams coming out again from Atlanta, Georgia.

And, with the latest project, the question was about doubling and tripling the capacity. On one hand, we don't have that much traffic unlike bigger CDNs obviously than us. But on the other hand, what our approach is, obviously, is adding more telecom providers.

But on the other hand, from software/hardware point of view, we're talking about specialized hardware. We're talking sometimes FPGA-based hardware which allows us to go beyond half terabit per one user. We talked about, like Peter was talking about, the power consumption. So we're talking about 120 watts maybe which will generate half a terabit of traffic; it's actually a pretty good number. And, this is what we're dealing with actually, and this is our focus. How cost-effectively from all points of view, not only, obviously, our point of view.

But, from the point of view of size, from the point of view of maintenance, from the point of view of how much engineers do you have to have to optimize a software-based solution versus when you have an FPGA solution, you can just plug and play. Well, this is our approach. Vertical scaling. As much as possible, vertical scaling, which requires, in our opinion, less support, less management, less troubleshooting. It's much easier.