Tutorial : Reduce oscam cpu (affected by cache-ex)

[tommys1]

the screenshot above is using not clocked q9450
just curious what is your i7 model nr exact Ubuntu Version and how many oscam instances or/ and other emu running at the same time?

sorry i missed this post..

Description: Ubuntu 12.10
Release: 12.10
Codename: quantal
processor: Core i7-2860QM

[TV_LICENCE]

hi!

I do not wish to reply in this tread any further ..... because it will upset someone ( or maybe even more users) ..... please do not ask me why and do not force me doing it

thanks

[tommys1]

hi!

I do not wish to reply in this tread any further ..... because it will upset someone ( or maybe even more users) ..... please do not ask me why and do not force me doing it

thanks

ok - but i thought we were sharing ideas and trying to improve oscam on this part of the forum?

[CapNCooK]

ok - but i thought we were sharing ideas and trying to improve oscam on this part of the forum?

We are. No worries. (-and no drama!)

[turbopower]

I`ve read in one cacheex exchange thread.

I have OSCAM 1.20-unstable_svn build #7983 with cyclepatch and high cpu patch.....

Did you have more information about this? Is there some significant difference with this patch?

[tommys1]

I`ve read in one cacheex exchange thread.



Did you have more information about this? Is there some significant difference with this patch?

i am not aware of this patch.

capncook, are you aware of this?

[CapNCooK]

i am not aware of this patch.

capncook, are you aware of this?

Yes. Patch is in SVN at this time.


See conclusions here: http://www.streamboard.tv/wbb2/threa...d=36453&page=4 (page 4, bottom threads)

OK - After extensive testing, i come to these conclusions;

The CPULoad patch;
- does improve the way oscam transports cache (flow)
- does not directly lower CPU in % on high load systems
- does spread load better across available threads
- supresses some logging to save more CPU cycles
- does re-accept old formatted packets.

--

Besides the effects of the patch, i learned the following;

- The exchange protocols are at least 'not very intelligent' by design.
- Partly therefore, oscam is not optimized of transporting mass ecm's from all peers to all peers
- High CPU loads (above 100%) is causing cache delay, causing less cache hits.
- High CPU loads are related to USERS in mode 3. Changing READERS have 0 effect on load.

Therefore i advise on high load systems to keep CPU below 100% !
(Check CPU load with top.. If its below 100%, do nothing. Its fine.)

If it is high..:

- Create a basic set of CAID's that you support, and apply these as filter to all users
for example: cacheex_ecm_filter = 0963
- Remove all existing cacheex_maxhops settings from all users. (or put to 10).
- Then, check cacheex stats to identify peers that give you highest hitrates
- Then, apply cacheex_maxhops = 2 to all users (to limit incoming cache very strict)
- Now check CPU with top (which will be low).
- Then, start to increase cacheex_maxhops ONLY on your top cachepeers at primetime, until cpu reaches avg. 100%.

After doing this, you will see that overall cache-hits will increase.
Mine is now 92% for 0963, and around 85% for 0100:6A with cpu between 40 and 90.

Hope this helps!

[turbopower]

I update oscam to 8022 and add cacheex_ecm_filter = to every account but my cpu is still over 100%

[CapNCooK]

With =2 ?

Then apply =1 to all..

Same procedure after..

Please read again..

Its not parameter cacheex_ecm_filter.., but its cacheex_maxhops.
Put that to = 2 on all users..

[vincent47291]

Please read again..

Its not parameter cacheex_ecm_filter.., but its cacheex_maxhops.
Put that to = 2 on all users..

That version doesn't do it for me! I tried 8024 with cacheex_maxhops set to 2 on both the users and the readers and my CPU time was through the roof! I reverted back to 7983 and everything was fine!