Caching In Part II
So what is edge caching? Edge caching is simply provision for a network content cache at the point where a local ISP network joins the Internet at large. It is not a new idea, pretty much every Web browser in use today supports HTTP proxy caching. the difference is scale. A HTTP proxy cache does not typically keep copies of video resources.
In recent years edge caching has been rather less fashionable than Napster style 'peer-2-peer'. P2P bypasses the need for the ISP to invest in cache infrastructure by conscripting end user machines as caches. This is good for the P2P provider but very bad for the ISP as the content will now travel over the most constrained part of the ISP's network multiple times.
The value of edge cachine is already known to companies like Akamai of course. But Akamai is a proprietary scheme. Google recently began work to build out a similar scheme and there will be many more as Internet video on demand becomes an increasingly bigger market.
So pity the poor ISP who is expected to provide space and power for all these boxes in their endpoints. If any economics student is looking for a thesis topic, try predicting which parties will benefit from this particular arrangment during the introductory phase and then again some years later once consumers have reliable ways of measuring network performance being delivered.
My rough model suggests that under the proprieatry cache model each party benefits at exactly the wrong time. In the short term, some ISPs may gain a modest revenue stream but in the longer term content is king.
Rather than waiting passively for the content distribution companies to come along with their boxes, a better strategy for the ISPs would be to develop a model that puts the edge cache under their control, allowing the ISP to determine the choice of hardware/software platform and which content content is cached.
The design of a network protocol for such a scheme may be left as an exercise for the (graduate) student. A discovery mechanism will be required (hint, SRV records in the reverse DNS) and some means of breaking content up into manageable chunks. And in the case of really popular content there will be a need for load balancing amongst local servers.
The rather more interesting issue is the security considerations that arise. Who gets to store content? Who gets to retreive it? When is content deleted? How are questions of copyright ownership decided?
