You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@ignite.apache.org by Peter <gr...@gmx.de> on 2018/11/26 15:32:50 UTC
Speed Expectations for Ignite Queue
Hello,
I'm currently trying Apache Ignite and love the concept.
I have created a simple example of what I'm trying to achieve
(producer&consumer):
https://github.com/karussell/igniteexample/blob/master/src/main/java/test/MyIgnite.java
My assumption is that the default settings are tuned to get a setup that
is okayish-fast and also without persistence and so comparing the speed
to LinkedBlockingQueue shouldn't be that unfair. My expectation here is
that ignite is 10-50x slower due to serialization overhead still
happening for localhost.
But it seems that ignite queue is 100-500 times slower. What am I doing
wrong here or did I simply have too optimistic expectations? Maybe
non-default settings like here
<http://apache-ignite-users.70518.x6.nabble.com/Impact-of-connectionsPerNode-and-pairedConnection-td18996.html>
would be better suited? See also MyIgniteSingle1.java where I also warm
up everything before measuring the speed. I also tried various different
JVM settings <https://apacheignite.readme.io/docs/jvm-and-system-tuning>
without success.
Kind Regards
Peter
PS: I cross-posted this a few days ago on SO:
https://stackoverflow.com/q/53454232/194609
Re: Speed Expectations for Ignite Queue
Posted by Ilya Kasnacheev <il...@gmail.com>.
Hello!
I think that you had too optimistic expectations. LinkedBlockingQueue is a
core Java data structure and it is very fast - it just writes a reference
to data structure in best case scenario.
Meanwhile, Ignite Queue will create several internal temporary objects
(such as distributed locks/futures) while processing addition or removal as
well as doing serialization.
Regards,
--
Ilya Kasnacheev
пн, 26 нояб. 2018 г. в 18:32, Peter <gr...@gmx.de>:
> Hello,
>
> I'm currently trying Apache Ignite and love the concept.
>
> I have created a simple example of what I'm trying to achieve
> (producer&consumer):
>
>
> https://github.com/karussell/igniteexample/blob/master/src/main/java/test/MyIgnite.java
>
> My assumption is that the default settings are tuned to get a setup that
> is okayish-fast and also without persistence and so comparing the speed to
> LinkedBlockingQueue shouldn't be that unfair. My expectation here is that
> ignite is 10-50x slower due to serialization overhead still happening for
> localhost.
>
> But it seems that ignite queue is 100-500 times slower. What am I doing
> wrong here or did I simply have too optimistic expectations? Maybe
> non-default settings like here
> <http://apache-ignite-users.70518.x6.nabble.com/Impact-of-connectionsPerNode-and-pairedConnection-td18996.html>
> would be better suited? See also MyIgniteSingle1.java where I also warm up
> everything before measuring the speed. I also tried various different JVM
> settings <https://apacheignite.readme.io/docs/jvm-and-system-tuning>
> without success.
>
> Kind Regards
> Peter
>
> PS: I cross-posted this a few days ago on SO:
> https://stackoverflow.com/q/53454232/194609
>
Re: Speed Expectations for Ignite Queue
Posted by Peter <gr...@gmx.de>.
To answer my own question: the speed seems to be expected and lays in
the same range like e.g. redis+redisson
Peter
Am 26.11.18 um 16:32 schrieb Peter:
>
> Hello,
>
> I'm currently trying Apache Ignite and love the concept.
>
> I have created a simple example of what I'm trying to achieve
> (producer&consumer):
>
> https://github.com/karussell/igniteexample/blob/master/src/main/java/test/MyIgnite.java
>
> My assumption is that the default settings are tuned to get a setup
> that is okayish-fast and also without persistence and so comparing the
> speed to LinkedBlockingQueue shouldn't be that unfair. My expectation
> here is that ignite is 10-50x slower due to serialization overhead
> still happening for localhost.
>
> But it seems that ignite queue is 100-500 times slower. What am I
> doing wrong here or did I simply have too optimistic expectations?
> Maybe non-default settings like here
> <http://apache-ignite-users.70518.x6.nabble.com/Impact-of-connectionsPerNode-and-pairedConnection-td18996.html>
> would be better suited? See also MyIgniteSingle1.java where I also
> warm up everything before measuring the speed. I also tried various
> different JVM settings
> <https://apacheignite.readme.io/docs/jvm-and-system-tuning> without
> success.
>
> Kind Regards
> Peter
>
> PS: I cross-posted this a few days ago on SO:
> https://stackoverflow.com/q/53454232/194609
>