[Spce-user] Server requirements

Matthew Ogden matthew at tenacit.net
Tue May 27 10:28:41 EDT 2014


Hi

You can also read
http://www.vmware.com/files/pdf/techpaper/voip-perf-vsphere5.pdf

Just as an fyi. Its not SPCE specific.

I think Jon points to HDD io because of call tear ups and tear downs (cdr,
and logging etc), but the io load is very small. We have a system in VMware
running 16 subscribers (PBXs), and averaging about 10 calls during business
hours most of the day. We also run queries to its API every 30 seconds for
each of the subscribers. We also are running a tailing packet capture all
the time using tshark on the same box. So we have two non-standard
activities going on.

Its averages about 7% CPU of 4 vCPUs (though I notice this is mostly from
our API queries),  the datastore IO is averaging 60iops during business
hours, 0.34mbps. Amount of active memory is 500MB

So just to highlight "one configuration out there", and do some theory,
maybe in practice it could go either way, but in clusters of virtualization,
it tends to scale up as the theory suggests until you reach 90%. And I'm
sure this would be the same if I just made it one larger system with more
ram, and more vCPU.

So our configuration on a 12 core system maxing at 90% CPU would be
something like 38 times these numbers, and the iops would be 2280 (just
inside my spec for our HDDs), and 12Mbps disk throughput (well well within
spec) and 19GB of RAM , I assume the Iops on a single system would be less
than on multiple different systems.

And that would that would mean with us running tshark all the time and
querying the API every 30 seconds we could scale to 380 calls, with 608
subscribers... though the subscribers could probably much higher, also I
think  your average number of active channels as a percentage lowers down
the more subscribers to have compared just a few subscribers.

I my answer along with others  is helpful.

Kind regards

-----Original Message-----
From: spce-user-bounces at lists.sipwise.com
[mailto:spce-user-bounces at lists.sipwise.com] On Behalf Of Jon Bonilla
(Manwe)
Sent: 27 May 2014 10:40 AM
Cc: spce-user at lists.sipwise.com
Subject: Re: [Spce-user] Server requirements

El Mon, 26 May 2014 13:10:28 -0400
"Santiago Albanese" <santiago.albanese at gmail.com> escribió:

> Hello:
>
> I am looking for recommendations regarding server specs. I have a new
> Dell R520, 96MB RAM, several TB of disc space on chassis and dual
> Intel Xeon
> E5-2430 v2 2.50GHz, 15M Cache, 7.2GT/s QPI, Turbo, 6 cores each, dual
> internal SD cards. And a Dell R620 16MB RAM, 160GB SSD Drives, One
> Intel Xeon E5-2620 2.00GHz, 15M Cache, 7.2GT/s QPI, Turbo, 6 cores and
> Dual internal SD cards. Can I run several instances (or one) of SPCE
> in a vmware environment on the R520 or is it better to have the R620
> dedicated for SPCE?
> I can figure out that dedicated is better than shared, but looking for
> experiences on running it virtualized and resources it needs. I assume
> the
> R520 has plenty of power to "dedicate" to the SPCE via VMware.
>
> Any thoughts would greatly be appreciated.
>
> Santiago
>


Hi Santiago

It depends on the load you will have on the system. Are we talking about 50
concurrent calls or 500?

Id's say as long as your HD IO is good enough you won't have problems
running it virtualized. Although if you hit high load I would use dedicated
hardware and take care about IO.

In my experience RAM is the most important thing to have in the SPCE. Then,
HDD IO (raid 10 is a good idea IMHO). CPU is not that important or
intensive. If running in virtual environments take care if HDD speed and
network card interruptions.


cheers,

Jon


_______________________________________________
Spce-user mailing list
Spce-user at lists.sipwise.com
http://lists.sipwise.com/listinfo/spce-user




More information about the Spce-user mailing list