Yeah, initial tests show improvements just by switching to async get. On Tue, 3 Sep 2019 at 11:06, John Smith <java.dev....@gmail.com> wrote:
> Actually, I looked closer at my code. Cannot use getAll() and the Queue is > single partition, so can't use multiple threads, the application is per > event and we have tested it up to 3K events per second not including the > cache. > > But also the application a fully async using vertx.io. Looking closer at > the code cache.get() is a blocking call which vertx.io doesn't like to > much. > > So I will switch to IgniteFuture. I'm pretty confident that will help > allot. > > > > On Mon, 2 Sep 2019 at 07:13, Ilya Kasnacheev <ilya.kasnach...@gmail.com> > wrote: > >> Hello! >> >> Can you switch to getAll()? Alterntively, try to use more than 1 thread. >> >> Regards, >> -- >> Ilya Kasnacheev >> >> >> пт, 30 авг. 2019 г. в 19:11, John Smith <java.dev....@gmail.com>: >> >>> Actually some more details. >>> >>> I have a partitioned cache with about 4Million records over 3 nodes. >>> When I do a get from the REST API I can hit about 4000 Get per second. >>> >>> From my application I can only achieve about 200 Get per second. My >>> application is single threaded consumer consuming from single partition >>> topic and then does a GET per kafka record. I know as a single consumer >>> thread the application without cache can handle give or take 2000. So if >>> figure with a bit of tuning I can get up to 1000 Gets. >>> >>> Also using async GET cache. >>> >>> On Fri, 30 Aug 2019 at 12:00, John Smith <java.dev....@gmail.com> wrote: >>> >>>> Sorry the regular full mesh client. Maybe some threadPoolSizes >>>> >>>> On Fri, 30 Aug 2019 at 11:23, Alexandr Shapkin <lexw...@gmail.com> >>>> wrote: >>>> >>>>> Hello, >>>>> >>>>> >>>>> >>>>> Not a thin-client tuning in general, but you can check a serialization >>>>> settings in order to make sure you do not use a default one. >>>>> >>>>> >>>>> >>>>> *From: *John Smith <java.dev....@gmail.com> >>>>> *Sent: *Friday, August 30, 2019 6:04 PM >>>>> *To: *user@ignite.apache.org >>>>> *Subject: *Tuning client performance. >>>>> >>>>> >>>>> >>>>> Hi, is there any specific client settings we can set to tune the >>>>> client performance, maybe some thread pools or any stuff like that? >>>>> >>>>> >>>>> >>>>