By Art Reisman, CTO, http://www.netequalizer.com
As the Internet continues to grow with higher home user speeds available from Tier 1 providers, video sites such as YouTube , Netflix, and others are taking advantage of these fatter pipes. However, unlike the peer-to-peer traffic of several years ago (which seems to be abating), These videos don’t face the veil of copyright scrutiny cast upon p2p which caused most p2p users to back off. They are here to stay, and any ISP currently offering high speed Internet will need to accommodate the subsequent rising demand.
How should a Tier2 or Tier3 provider size their overall trunk to insure smooth video at all times for all users?
From measurements done in our NetEqualizer laboratories, a normal quality video stream requires around 350kbs bandwidth sustained over its life span to insure there are no breaks or interruptions. Newer higher definition videos may run at even higher speeds.
A typical rural wireless WISP will have contention ratios of about 300 users per 10-megabit link. This seems to be the ratio point where a small businesses can turn a profit. Given this contention ratio, if 30 customers simultaneously watch YouTube, the link will be exhausted and all 300 customers will be experience protracted periods of poor service.
Even though it is theoretically possible to support 30 subscribers on a 10 megabit , it would only be possible if the remaining 280 subscribers were idle. In reality the trunk will become saturated with perhaps 10 to 15 active video streams, as obviously the remaining 280 users are not idle. Given this realistic scenario, is it reasonable for an ISP with 10 megabits and 300 subscribers to tout they support video ?
As of late 2007 about 10 percent of Internet traffic was attributed to video. It is safe to safe to assume that number is higher now (Jan 2009). Using the 2007 number, 10 percent of 300 subscribers would yield on average 30 video streams, but that is not a fair number, because the 10 percent of people using video, would only apply to the subscribers who are actively on line, and not all 300. To be fair, we’ll assume 150 of 300 subscribers are online during peak times. The calculation now yields an estimated 15 users doing video at one time, which is right on our upper limit of smooth service for a 10 megabit link, any more and something has to give.
The moral of this story so far is, you should be cautious before promoting unlimited video support with contention ratios of 30 subscribers to 1 megabit. The good news is, most rural providers are not competing in metro areas, hence customers will have to make do with what they have. In areas more intense competition for customers where video support might make a difference, our recommendation is that you will need to have a ratio closer to 20 subscribers to 1 megabit, and you still may have peak outages.
One trick you can use to support Video with limited Internet resources.
We have previously been on record as not being a supporter of Caching to increase Internet speed, well it is time to back track on that. We are now seeing results that Caching can be a big boost in speeding up popular YouTue videos. Caching and video tend to work well together as consumers tend to flock a small subset of the popular videos. The downside is your local caching server will only be able to archive a subset of the content on the master YouTube servers but this should be enough to give the appearance of pretty good video.
In the end there is no substitute for having a big fat pipe with enough room to run video, we’ll just have to wait and see if the market can support this expense.