Negative lookup is slow which makes me think that swap or persistent store is configured.
Negative lookup flow maybe the following: local lookup (fail) -> send request to affinity node -> memory lookup on affinity node (fail) -> swap lookup (?) -> store lookup(?) Yury, can you please share you config? --Yakov 2015-10-28 12:07 GMT+03:00 Dmitriy Setrakyan <[email protected]>: > > > On Tue, Oct 27, 2015 at 4:11 PM, Yury Lazouski <[email protected]> > wrote: > >> All data is available locally which is proved by fast lookups or existing >> data in cache, but also we have different scenario - time to time we are >> looking for values that don't exist in cache at all (where lookup returns >> null) and these lookups cost (cost increases with addition of nodes, at >> least 1 or 2 nodes makes difference) >> > > Where are you loading the data from? I am still confused why the data is > absent on one node, but present on the other, when using a Replicated cache. > > >> >> >> >> -- >> View this message in context: >> http://apache-ignite-users.70518.x6.nabble.com/Performance-issue-with-REPLICATED-cache-tp1710p1730.html >> Sent from the Apache Ignite Users mailing list archive at Nabble.com. >> > >
