Open GoogleCodeExporter opened 9 years ago
Hi Andrei,
do you mind making your tests again by using AB (Apache Benchmarking Tool,
http://httpd.apache.org/docs/2.3/programs/ab.html)?
Original comment by sergio.b...@gmail.com
on 11 Mar 2011 at 9:52
Original comment by sergio.b...@gmail.com
on 12 Mar 2011 at 9:46
sorry I used jmeter, is it ok?
######################################3
Sergio:
I tested on the same box:
* with jmeter 2.4
* empty terrastore (all buckets removed, fresh restart, java warmed
up)
* single thread jmeter thread group
* on apache: request http://127.0.0.1
* on terra server: request http://127.0.0.1:8000
* total run size: 100K requests;
throughput results:
apache: 6500 / sec
terra server: 490 / sec
this is consistent with with my previous test;
thank you
Andrei.
On Mar 11, 4:50 am, Sergio Bossa <sergio.bo...@gmail.com> wrote:
Original comment by Andrei.Pozolotin
on 14 Mar 2011 at 7:53
Sure, JMeter is okay.
But again, Terrastore is meant to be multi-threaded, so you should actually run
your tests with multiple threads.
Original comment by sergio.b...@gmail.com
on 15 Mar 2011 at 5:25
yes, I get the "multi-threaded" part :-)
what I do not get is why it is 10 times slow on empty request for this simple
use case?
I assume in this case terrastore:
* does not introduce cluster locks
* bucket list (map with buckets) is already cached
* everything is read-only
then it should be "instant", right? :)
like I mentioned - with terrcotta toolkit and native terrcotta maps
I get close to apache speed;
well I guess this is just a request to take a fresh look on overall terrastore
pipline;
and if piplene is pluggable, may be try to plug in some other guys:
http://www.restlet.org/
etc
Original comment by Andrei.Pozolotin
on 15 Mar 2011 at 8:35
Agreed: we should give other options, such as Restlet or Jersey, a try.
Original comment by sergio.b...@gmail.com
on 16 Mar 2011 at 1:14
Original issue reported on code.google.com by
Andrei.Pozolotin
on 10 Mar 2011 at 4:18