John Linton
Exetel put in place Deep Packet Inspection processes almost a year ago to control the too rapid growth of the use of P2P usage in peak demand times. This, after an initial 'teething period', has worked very well in increasing the overall use of the bandwidth deployed by Exetel while making little/no impact on the overwhelming majority of Exetel's DSL customers. In discussions with the supplier of the equipment we selected, and with suppliers of other DPI equipment and software we considered, it seems that Exetel's pioneering work has been copied by almost every other ISP in Australia all of whom have now introduced similar processes (and the several who introduced such equipment and processes before Exetel have continued to upgrade and use the equipment they purchased). The few customers who left Exetel because we were spreading their P2P downloads over a longer period ended up with ISPs who did exactly the same thing only, in all probability, not as well as Exetel did it.
During the period we were deciding on the P2P control solution we eventually selected we considered several other alternatives to the control type of solution including three caching of specifically P2P traffic solutions then just appearing on the market. The attraction of a caching rather than a control solution was, of course, that it delivered all P2P data at the time of request but reduced the use of ingress/egress bandwidth similarly to a P2P control solution - it's major drawback over the control solution being that it didn't reduce the bandwidth use between the customer and Exetel which, with the introduction of ADSL2 from Optus and Powertel, was more expensive than from Telstra. Apart from that financial disadvantage, the technology (not cache technology per se but the P2P specific caching technology) was very new and the developers of those services had very little in the way of reference users and what they did have weren't really helpful in Exetel's assessments.
Following Steve's visit to InterOp earlier this year we re-initiated our investigations of caching P2P based on two developments - the fact that the technology was a year older and had developed quite considerably over that period and the fact that there were now at least two reasonable reference ISPs that were closer to Exetel's 'profile'. I, personally, still saw the inability to reduce the customer to Exetel bandwidth as a major drawback, particularly as the "sales pitch" had changed over the 12 months from emphasising savings in bandwidth to increasing P2P user satisfaction by actually making MORE bandwidth available to P2P users! It became clearer in subsequent discussions that, while it appeared to be a major negative, at least to people like me, there was some sensible logic in what was now being proposed.
The logic centred on the increase in legitimate streaming data based on P2P protocols to provide a multiplicity of legal downloads of movies, TV shows and more sophisticated delivery of major software 'titles'.
While I'm far from convinced that the time for an ISP to promote the use of P2P as the download protocol of choice is now - I can see the logic of doing that if Disney, Sony, Universal, Microsoft etc actually do end up deciding to use it. As I pointed out to the enthusiastic P2P caching person who visited us yesterday, Akamai semed to have built their business on persuading the major copyright holders to use Akamai around the world to protect their copyright and it would take some persuasion, not to mention contract re-negotiation, to change what is currently in place.
However, I'm intrigued with the concept of moving from 'controlling' the use of P2P to 'promoting' the use of P2P and of doing so as a financial benefit to Exetel and Exetel's users. We will look at the developments of this concept over the next three to four months with a great deal of interest as our current view is that the use of caching would not only not produce a financial advantage it would pose a financial burden. Having said that we have agreed to continue the investigations of how to use the proposed streaming functionality with the objective of implementing them early in 2008 if the claims currently being made can be substantiated.