[cisco-voip] CUCM on 3rd party hardware (CPU question)

Mon Apr 1 10:48:12 EDT 2013

You can also post in the collab user group.  I believe the PM that owns this bit of policy is active there and would likely be able to give you a specific reason why they came up with the speed they did for that processor.

http://www.cisco.com/en/US/products/ps6884/products_tech_note09186a0080bf23f5.shtml is also an appropriate read here.  You are solidly in the “If it works, is it allowed by the vendor’s support policy rules?” category, and need to be very aware of the support implications that come with it.  

You may be perfectly fine, but in the event you have any issues you need to be aware of what's going to happen.  This is a simple risk analysis, and you should be comfortable with whatever decision you make now (and document it in case you have to justify it at a later date).   If you do proceed with this cpu at least make a point to under-utilize your servers.  

As an aside the 1000-user OVA is only for BE6K (and the TRCs it is based off + standalone ELM).  You can't and shouldn't use it outside of those situations.  

Finally, don't rely on CPU reservations for justifying how much cpu a UC app may need.  We don't allow CPU oversubscription for a reason (so the app can get the full core if need be), and any reservation an app may or may not set is for minimum processing power.  Things don't tend to fall apart in an idle server.  Where the processing really counts is under heavy load, which is most commonly associated with some other unexpected event (failover, call loop, etc).  That isn't the time when you want to start seeing process cores because the system can't keep up with the incoming rate of signals.  Those types of cascading failures are generally when exec's start calling each other, and everyone starts talking about root cause, design assurance, etc.

-Ryan

On Mar 29, 2013, at 2:41 PM, "Madziarczyk, Jonathan" <JMad at cityofevanston.org> wrote:

Thanks Matthew,

Yeah, I’ll need to whitebox at some point, once I can get some media.

From the CPU side, it’s almost $3k to upgrade a single server that isn’t even 3 months old.

Ah well, I’ll try at the SE/AM level and see what I get.

Jon

From: Matthew Loraditch [mailto:MLoraditch at heliontechnologies.com] 
Sent: Thursday, March 28, 2013 3:50 PM
To: Madziarczyk, Jonathan; cisco-voip at puck.nether.net
Subject: RE: CUCM on 3rd party hardware (CPU question)

Those specs are carved in proverbial stone. You can see in the not supported column your scenario is specifically denied. While I believe you are right about actual performance, the test teams only have so many resources and there have to be some rules. I think it’s pretty fair as is right now, they only thing they really care about any more is the CPU you can build a white box at this point if you want. You still need enough RAM and such but no requirements as to brand, speed, etc.

You can certainly talk to your AM and SE and provide the feedback, but I’m 99% sure there is no such thing as a dispensation. Why don’t you just get new CPUs if everything else is ok with your servers?

Matthew G. Loraditch – CCNP-Voice, CCNA, CCDA

1965 Greenspring Drive
Timonium, MD 21093

voice. 410.252.8830
fax.  410.252.9284    

Twitter  |  Facebook  | Website  |  Email Support

From: cisco-voip-bounces at puck.nether.net [mailto:cisco-voip-bounces at puck.nether.net] On Behalf Of Madziarczyk, Jonathan
Sent: Thursday, March 28, 2013 4:28 PM
To: cisco-voip at puck.nether.net
Subject: [cisco-voip] CUCM on 3rd party hardware (CPU question)

So I’m trying to figure this one out.  Maybe some Cisco gurus can chime in?

I started by going off of this document:http://docwiki.cisco.com/wiki/UC_Virtualization_Supported_Hardware#Processors_.2F_CPUs
My brand new Dells have Intel Xeon E5-2660s in them which qualifies, but they’re running at 2.2GHz instead of the requested 2.5GHz. 

However when I look at this: http://docwiki.cisco.com/wiki/OVA_Template_Details_for_Unified_CM_Release_9.0
I see that both the 1000 and 2500 user servers are based off of the E5500 series chips which is significantly slower than my processors, as based on the following: http://ark.intel.com/compare/37104,37096,64584 

Technically, the 1000user server model doesn’t even qualify as a valid hardware config based off of the first docuwiki link, and both models are using 4yr old CPUs while I’m trying to use a <1yr old cpu that’s 2 generations ahead of Cisco’s recommendations (2x threads, 2x cache, 2x bus speed, 2x memory bandwidth, more memory channels, faster Max Turbo Freq, etc).  And secondly the reservation is only for 800MHz, which means if there’s any contention, having a 2.5GHz chip means nothing, right?

I get that Cisco is trying to pad their stats to eliminate any sort of a hardware issue, but from my perspective using a brand new server I already own vs. spending another $8k.  Especially, when I believe existing will run even better than the OVA requirements.

Obviously I want to be TAC supported on the software, do I need to ask for special dispensation to run on existing hardware (would that even have a chance)? 

Thoughts?
_______________________________________________
cisco-voip mailing list
cisco-voip at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-voip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/cisco-voip/attachments/20130401/4b0ca715/attachment.html>