GoogleCodeExporter
commented
9 years ago related IRC discussion:
<yulian_o> in 3.0 number of resources is usefull , its also makes the
difference , i have sent mail to some team members with wireshark from mms that
started with 50K of endpoints , it did not added any load at all since most of
them where idle
[11:46] <ivelin> right, some info about the schedule could be useful so
sysadmins can plan their hardware/MMS resources based on anticipated subscriber
load
[11:46] <tommyq> oh that's what that was!
[11:46] <tommyq> :)
[11:46] <tommyq> I was wondering how you got that many endpoints on one box
[11:47] <yulian_o> simple answer : resources are global in 3.0 so endpoint is
using less then 100 bytes , may be a bit more , 50K is several megabytes
[11:47] <yulian_o> 50K rtp connections is much more difficult to gain :)
[11:47] <tommyq> :)
[11:48] <tommyq> understood now it all makes sense
[11:48] <yulian_o> in previous release all endpoints came with predefined
resources so you could not start with such huge number of endpoints , it would
lead to 1 million local + rtp connections preloaded
[11:48] <jean> ah I see now :-)
[11:48] <ivelin> why not start endpoints on demand?
[11:49] <ivelin> or use a flex pool - start 100 more when almost 100 are being
used
[11:49] <yulian_o> its the issue of naming basically , i though about that ,
the problem is in naming of endpoints
[11:49] <yulian_o> you have mobicents/ivr/1...10 for example loaded
[11:49] <yulian_o> its 10 endpoints
[11:50] <yulian_o> since the endpoint name is textual you can not add them
freely , theoretically it can be mobicents/ivr/a...h
[11:51] <yulian_o> but if limititations on name will be set , last node should
be numeric it will make mgcp search faster and also will allow increasing
number of endpoints on runtime
[11:51] <yulian_o> rtp connection for example does not have such problem
because it has numeric id , so you have long number and can expand it as you
wish
[11:53] <ivelin> why not use long numbers for end-poins?
[11:53] <yulian_o> because endpoint names are numeric
[11:54] <yulian_o> sorry textual
[11:54] <yulian_o> http://tools.ietf.org/html/rfc2705#section-2.1.2
[11:55] <yulian_o> if we agree that last node will be numeric we can allow it
growing in case its not ss7 endpoint
[11:55] <ivelin> lets take this offlne
[11:56] <jean> who defines the names of the endpoint identifiers ?
[11:56] <ivelin> I'd like to understand it better
[11:56] <jean> us ?
[11:56] <jean> if so we can mandate that the endpoint names textual value is a
numeric one
[11:56] <jean> that can be used internally as numeric
[11:56] <yulian_o> we define , and can mandate that last node is numeric one
[11:57] <charles_r> .
[11:57] == charles_r [c9c1fa7a@gateway/web/freenode/ip.201.193.250.122] has
quit [Quit: Page closed]
[11:57] <yulian_o> then instead of defining /mobicents/ivr/1...10 in config
[11:57] <yulian_o> we can make /mobicents/ivr/ and add additional param number
of endpoints
[11:58] == charles_r [c9c1fa7a@gateway/web/freenode/ip.201.193.250.122] has
joined #mobicents
[11:59] <jean> cool
[12:00] <yulian_o> so should i open the issue?
[12:00] <ivelin> yes, please
[12:00] <yulian_o> ok
[12:00] <ivelin> #7 Productization
[12:00] <slegrik1> I am continuing on jss7 Telscale release script for
Telscale-jss7-6.1.3.GA and this week there was only a little progress on my
side, I was able to prepare an initial
[12:00] <slegrik1> release script, but nothing more...
[12:01] <slegrik1> So basically the target for this week is moving to next one.
[12:01] <yulian_o>
https://code.google.com/p/mediaserver/issues/detail?id=99&thanks=99&ts=135231128
2
Original comment by ivelin.a...@gmail.com on 7 Nov 2012 at 6:55
Original issue reported on code.google.com by
oifa.yul...@gmail.comon 7 Nov 2012 at 6:01