A fair number of online games that are big use P2P, notably the Blizzard games (Warcraft, Starcraft) to push out large updates.
This is mostly done not for performance reasons but for economic reasons: shifting the cost for delivering large (often VERY large) patches or even entire games (eg, Starcraft is several GB) from the content provider to the recipient's ISP. And economic constraints are what makes P2P interesting. Localization helps in such a framework, by deloading the access link network both up and down, but does nothing to deload the last mile network which for some networks (eg, DOCSIS) can be the big bottleneck/problem on P2P traffic, which means that although localization would improve the P2P traffic, the cost to the ISP might still be much greater than non-P2P if the local-access uplink costs more than the general access downlink. In fact, some ISPs would benefit from configuring the local Alto service to AVOID local seeds to save money, if Cost(local uplink) > (Cost(access uplink) + Cost(access downlink)). On Sep 30, 2010, at 8:02 AM, Rimac, Ivica (Ivica) wrote: > I agree with the huge number of Wikipedia users, but Victor’s point makes > sense. The demand for a service like alto depends on the pain the generated > traffic causes to the ISPs—why bother if you can squeeze out an improvement > of lower digit percent (the voice of devil’s advocate in me ;) )? > > A related question, does anybody know of Akamai’s position for alto-like > services for their NetSession solution? I am not aware of how prominent this > is anyway though I read somewhere that it is being used for distribution of > games. _______________________________________________ alto mailing list [email protected] https://www.ietf.org/mailman/listinfo/alto
