On Wednesday 26 April 2006 13:59, Ian Clarke wrote: > On 26 Apr 2006, at 10:57, Matthew Toseland wrote: > > On Wed, Apr 26, 2006 at 09:49:17AM -0700, Ian Clarke wrote: > >> I think when we have applications that explicitly rely on requesting > >> keys that probably haven't been inserted yet, the proportion of > >> successful requests is not a good measure of the network's > >> effectiveness. > >> > >> A good measure of the network's effectiveness is to manually insert > >> keys at one node, and request those keys from another node, > >> preferably one that is as far as possible from the first in the > >> network topology, while monitoring the success rate. > > > > That is probably true. But there needs to be a time lag to establish > > whether there is a problem with the data not keeping up with the > > location swaps. > > A good test will insert a bunch of CHKs, and then request them over a > period of time, say, one an hour for a week. That way, if > retrievability decreases with time, we will see the extent of this > problem.
Location swap had me worried. Since the alpha started I have kept 10-15 connections active. My node's location has stayed within a range between 0.12 and 0.99. If other nodes are behaving the same way we are probably in pretty good shape. One test we could do would be to try to retieve old version of sites. If they are still getable its a good indication that routing is working and location drift is slow enough not to be a big problem. From personal experience, most of the freesites are pretty dependable. Ed
