Archived
Indiscoverable performance issue. Help please.
Scene:
env:windows server 2008 r2. ice 3.6.1
role:
testclient: it will call server1's m1。
ice server1:with one servant which have one method named m1, which will call server2's m2。
ice server2:with one servant which have one method named m2, which will be called by server1's m1。
both ice server1 and server2 has communicator with 200 server threads max and 200 client thread max。 m1 and m2 do little things, exec time is about 1ms. testclient have a parameter which can change the concurrent call threads count to m1. And testclient's thread is created by windows api, not ice thread. testclient / server1/ server2 are deploy with different mathine. test and result: testclient thread count requests count(rough) 10 450000 100 600000 200 720000 400 600000 800 50000 (why???) And then, I tried to trun server1's server thread pool to 1000, and testclient thread count 800, then the requests count up to 700000 again. I want to know: 1: Why the testclient concurrent count can effect the response time seriously? 2: Why everything goes ok when I tuned the server1's server threadpool to larger then testclient thread count?
Comments
-
the requests count are result in 20 seconds
0 -
Yesterday, we tried the test again, with a independent thread pool handling requests, the ICE thread only repackage the request and throw it into our thread pool.
Under such circumstance, the problem disappeared.
So, we guess that the problem may be related with the IOCP event handling of ICE server thread pool.0 -
Is there anyone can help me?
0 -
Hi MingYi,
These forums are primarily for simple questions from developers new to Ice. For more advanced questions, you should consider subscribing to our Priority Support, and then send us a support request. See https://zeroc.com/support.
Best regards,
Bernard0