0 Members and 1 Guest are viewing this topic.
the peer caches are broken down in to 10 for each region, then by version as well... you'll notice the 352 hostnames covers 3.52-3.54 for some reasoneach version has 6 regions, and each of those regions then has 10 hostnames in it and it will pick a random number for thateach request to the peer cache returns 12 the details of primary users, because if one fails it prevents having to return to the peer cache to get further details... although primary users once you have one connection it will use the data on that connection to locate primaries rather than the peer cache
1. i haven't looked at which numbers are which region so can't answer that however the idea of different peer caches for each reason only works if you only send results containing primaries set to use that cache... which of course means that the cache has to return results from primaries that have "checked in" with it... that is how frontcode did it, and that is why an average of about 50% of the results from the froncode peer caches failed...