Whenever I see latencies which are roughly multiples of 40ms it screams to me 'nagle algorithm'. I have seen this so often now, that the first thing I check is, if the TCP_NODELAY option is set on the TCP socket on both ends.
Am Freitag, den 25.02.2011, 08:14 -0600 schrieb Jeremiah Peschka: > Here's where the infrastructure guy in me kicks in: > > > What kind of drives are you using in your machines? What kind of NICs > do you have? Are the switches and NICs all forced to full duplex or > are they set to auto-configure? > > > Theoretically, you should be seeing performance much faster than 80ms > on GETs, or so I'd think. > > -- > Jeremiah Peschka > Microsoft SQL Server MVP > MCITP: Database Developer, DBA > > > On Thursday, February 24, 2011 at 10:12 PM, Wilson MacGyver wrote: > > > did you change the size of the thread pool? on the default setting > > of > > 64, I've noticed > > some high startup time with the GET. > > > > On Wed, Feb 23, 2011 at 3:08 PM, Bob Feldbauer > > <[email protected]> wrote: > > > After importing data into Riak, I'm in a basically 100% GET > > > scenario. I > > > haven't done much load testing other than throwing live traffic at > > > it, but > > > with that I've seen ~150 concurrent connections. > > > > > > I'm using Jetty app servers (Riak Java client) which are using > > > protobufs to > > > hit an HAProxy server, which load balances my 7 Riak nodes. I > > > really haven't > > > done anything special, other than using protobufs and fronting > > > Riak with > > > HAProxy. > > > > > > Other than that, I would just say "use fast disks" and/or add > > > nodes as > > > needed to hit your capacity targets -- because with that traffic > > > and 7 nodes > > > with 1 1TB disk per node, I'm seeing ~0.5-<1.0 load average on > > > Riak nodes as > > > iowait. > > > > > > Feel free to hop on the #riak IRC channel and I'd be happy to try > > > to answer > > > any other questions about our setup. > > > > > > - Bob Feldbauer > > > > > > > ---------- Forwarded message ---------- > > > > From: Wilson MacGyver<[email protected]> > > > > Date: Tue, Feb 22, 2011 at 7:15 PM > > > > Subject: tuning riak for 100s of concurrent connections > > > > To: riak-users Users<[email protected]> > > > > > > > > > > > > Are there any guidelines/howtos on tuning riak nodes for 100s > > > > (200-500) of concurrent connections > > > > which are 99.9% HTTP GET? > > > > > > > > Thanks, > > > > > > > > -- > > > > Omnem crede diem tibi diluxisse supremum. > > > > > > > > _______________________________________________ > > > > riak-users mailing list > > > > [email protected] > > > > http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com > > > > > > > > > > > > -- > > Omnem crede diem tibi diluxisse supremum. > > > > _______________________________________________ > > riak-users mailing list > > [email protected] > > http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com > > > > > _______________________________________________ > riak-users mailing list > [email protected] > http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com _______________________________________________ riak-users mailing list [email protected] http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
