John Linton As we prepare to install and test the new IP caching service I've been more closely following the 'discussions' in the USA concerning the handling of various traffic types (not that I play any major part in the managing of Exetel's network other than agreeing to pay for the constant upgrades).
When we first began to investigate how we could improve the speeds of various traffic types I was constantly reminded that from the period 1995 until 2006 I had always been told that caching could never be justified because the cost of bandwidth fell faster than the cost of disk space and disk controllers.
However the web is always a great way of finding out what almost everyone s doing in terms of technology and networking and we became more and more aware, and therefore interested, in the impacts of 'new' traffic types on 'old' network provisioning "policies". Apart from 'P2P' (irrespective of what any end user may think is 100% described by 100% of network managers as of major concern and impact) there was a great deal of discussion about the rapid growth of streaming video traffic and how it could be optimised without 'breaking' the network provisioning policies that covered other types of traffic.
At Exetel we have watched the slow increase (in percentage terms) of the use of streaming video which has continued to accelerate as YouTube/MySpace type facilities have become better known and more widely used. Today those two sites account for a little over 5% of the total traffic on our approaching 2 gb of bandwidth - an increase from less than 1% less than 18 months ago when we first deployed the ability to analyse traffic types by signature.
With the growing end user interest in IPTV (from remote sources rather than locally delivered from DSLAMs) and also the growing use of legitimate (paid for) movies via IP every comment from network managers in the USA is suggesting that video streaming will move to around 20% of total bandwidth use within the next 24 months - something I wouldn't have thought possible less than 12 months ago (when Exetel's streaming video delivery was around 1%).
Streaming video, even more than P2P in most situations, is a major 'bandwidth hog' and, far more than P2P, totally 'unforgiving' in terms of delivery acceptability.
It became obvious that in deciding on the caching solution Exetel would deploy a very important characteristic would be the ability to handle streaming video. There were dozens, probably hundreds, of streaming video caching solutions but only very few, actually none in production, for integrating IP and streaming video caching in the one 'solutuion'.
The solution we eventually selected didn't have this functinality when we made the decision to purchase it but it was a 'planned' (and I have a lot of trouble with that word) for availability around the time we expected to trial the devices. I was pleasantly surprised to see that the beta version of this software is available for the trial and it'll be very inmportant to ensure it works as effectively as we anticipate.
Having wresteled with P2P implications for the past two years the need to deal with streaming video is not something I look forward to but what concerns me more is what is the next 'network breaking' application that will loom up in the near future?