cancel
Showing results for 
Search instead for 
Did you mean: 

Thread/Process Limit on media server

Dr_Strangelove
Level 3
Hi,
 
Is there a thread or process limit on a NBU 6 MP4 media server ? I was wondering if a NBU media server running on a large box, such as a SUN 4900,  would hit software constraints before hardware constraints. I am checking out the possibility of putting in a 4900 media server, into a large environment, with a large amount of IP bandwidth coming in, say 60Gb and maybe 40Gb going out the back to the backup media infrastructure. I would think however that NBU would be the bottleneck in terms of the number of threads it could support before the hardware got maxed out.
 
Is anyone running NBU on big media servers ?
 
Thanks for any help.
5 REPLIES 5

DavidParker
Level 6
One of my environments has a fairly large crop to manage (almost 700 clients).
We did find some problems with the OS running out of available threads due to the nightly schedules kicking off like 2000 jobs at once.  NBU seemed fine with it, but Solaris didnt.

I don't have all the details on the problem/solution though, but thought I'd mention it.

Dr_Strangelove
Level 3
David.
 
What version of Solaris were you running ?

DavidParker
Level 6
I'm pretty sure it's Solaris 10.

Darren_Dunham
Level 6
There is no specific thread limit in Solaris 10 that you're likely to run into on real hardware.  At some point, you'll run out of some resource (likely memory) to create a new process or thread.    You can run out of process table entries for new processes, but again you're likely to run out of memory long before you run out of process slots.  That doesn't mean that it's easy to determine *where* the resource issue lies when it gets hit, though.

http://access1.sun.com/techarticles/limit.html is interesting and talks about differences in the old thread library (pre-Solaris 8) and the current one (used exclusively in Solaris 9 and later).

--
Darren

Dr_Strangelove
Level 3
Thanks for the info.
 
The box we are looking at can take 256GB of memory so we might be able to nail that resource as not being the bottleneck.