search

noushadali / mediaserver

Automatically exported from code.google.com/p/mediaserver
0 stars 0 forks source link

MGCP Client does not handles all requests under high load #94

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago 
Following :
https://groups.google.com/forum/?fromgroups=#!topic/mobicents-public/XKc2PqUyDd8

Example 
Request 830662670
First request sent at 41.777176 , request id 12295
Response received at 41.790378 , request id 12341
Another request sent at 45.968471 , request id 12616

That means that response is actually received by server but was not parsed by 
mgcp stack for some reason.

Original issue reported on code.google.com by oifa.yul...@gmail.com on 31 Oct 2012 at 9:38

GoogleCodeExporter commented 9 years ago 
Found an issue
the problem is that packet representation does not copies the data for receiver 
buffer , therefore causing reusing of same buffer over all messages received in 
20ms period.
This can lead to issue that message data is actually overriten and other 
message may disappear.
Also need to change how sleep time is calculated depending on delay

Original comment by oifa.yul...@gmail.com on 31 Oct 2012 at 5:15

GoogleCodeExporter commented 9 years ago 
Modified by revision  20f2137f80b0
Waiting for verification

Original comment by oifa.yul...@gmail.com on 31 Oct 2012 at 7:01

GoogleCodeExporter commented 9 years ago 
Does not helpes , there still exists some duplicate packets.
Need to separate reading thread and processing thread , possibly use several 
processing threads to increase power since currently uses one thread and 
therefore one cpu core

Original comment by oifa.yul...@gmail.com on 1 Nov 2012 at 11:14

GoogleCodeExporter commented 9 years ago 
Modified by revision dcd10023062c
Waiting for verification

Original comment by oifa.yul...@gmail.com on 2 Nov 2012 at 4:17

GoogleCodeExporter commented 9 years ago

Original comment by oifa.yul...@gmail.com on 2 Nov 2012 at 5:47

GoogleCodeExporter commented 9 years ago 
New Exception occured at mgcp client should set messagehandler per each parsing 
thread instead of single one.
Should optimize unique id generators at provider implement

Original comment by oifa.yul...@gmail.com on 5 Nov 2012 at 11:03

GoogleCodeExporter commented 9 years ago 
Updated by revision  592d8fe515aa

Original comment by oifa.yul...@gmail.com on 5 Nov 2012 at 12:41

GoogleCodeExporter commented 9 years ago 
Updated by revision  b14bd5e2a3fe
Waiting for load tests

Original comment by oifa.yul...@gmail.com on 2 Jan 2013 at 12:28

GoogleCodeExporter commented 9 years ago 
Updated by revision e1ce31d7269f

Original comment by oifa.yul...@gmail.com on 16 Jan 2013 at 9:54

GoogleCodeExporter commented 9 years ago

Original comment by oifa.yul...@gmail.com on 18 Jan 2013 at 9:15

GoogleCodeExporter commented 9 years ago 
Confirmed by restcomm team.
Got 1K concurrent calls ( Approximately 14K mgcp messages per second ). On 4 
cores ( 8 units ) EC2 instance

Original comment by oifa.yul...@gmail.com on 21 Jan 2013 at 1:32