On Tue, Oct 27, 2015 at 3:03 PM, Yury Lazouski <[email protected]> wrote:
> Hi, > haven't got a chance to create project yet, but will do however I moved > forward in investigation - I found out that positive cache lookups are > happening in 0-2 μs while negative (value is absent in cache) - >10 μs on > single node cache and ~150 μs!!! on 2-node cluster when data is read > locally!! Same positive lookups on 2-node cluster happen within the same > 0-2 > μs. > Yuri, 150 μs probably sounds reasonable for a network trip (even over a loopback socket). However, I am not sure why you even have network trips, considering that you are using REPLICATED caches. Shouldn’t all the data be available locally? > > Please let me know if it sounds like I miss something obvious otherwise I > will try to create simplified project to demonstrate this. > > Thanks, > Yuri > > > > -- > View this message in context: > http://apache-ignite-users.70518.x6.nabble.com/Performance-issue-with-REPLICATED-cache-tp1710p1728.html > Sent from the Apache Ignite Users mailing list archive at Nabble.com. >
