Out of interest how many devices are you looking to support in this design? Also is it a fresh install or an upgrade from an existing version of CM????

As Hailey mentioned, technically speaking there is no problem in mixing platforms. As long as you plan using the SRND to calculate the number of phones, TFTP, MOH, etc and use the right platform (7825/7835/7845), you should be fine.

AFAIK we have no problem supporting this, i believe someone from my team has a tech-tip on this on another of the Cisco Communities.

I've having a lot of things to do lately, so if someone could remind me on Monday, i can look it up. ;-)

HTH

java

if this helps, please rate

www.cisco.com/go/pdihelpdesk

Hi Java

thanks for the info... but as per usual I have a few more questions..

My client is looking to upgrade from their existing V4 to V7. Some of the V4 hardware is ok to use according to the compatibility matrix. They are also purchasing additional hardware to replace some of the older hardware.

I plan to use one of the new IBM's MCS7835I3 servers as my new Publisher, and I was going to use one of the existing HP MCS7835H1 servers as my dedicated TFTP Server. The concern I have is the comment on the SRND.

" Cisco recommends that you use the same hardware platform for the TFTP severs as used for the call processing subcribers, page 8-13"

Does this relate to just the 78xx version or the IBM / HP?....

So can I have my Pub as a new IBM 7835 and my TFTP as a HP 7835??? .... (below that I plan to have 2 CM groups, one pair of IBM7835's and one pair of HP 7835's)

Many thanks

Hi

My comment would be that the difference in hardware platforms is going to show up in certain scenarios:

1) Where you have a failover, and a large number of phones fail over to another server. If this server is of a lower spec, it may struggle with the failover and subsequent amount of additional traffic.

2) If there is a large difference in spec you might get poorer performance from a server if it's in a cluster where there is a large amount of intra-cluster signaling (i.e. a large count of phones or a system that is busy in terms of call counts and so on). Where there is a lot going on, there is more processing to do across all the servers regardless of how many phones are actually registered to the poorer spec server.

So I would ensure that your subscribers are pairs/triplets of even-performance boxes within the CallManager groups and keep your redundancy model a nice, simple predictable one.

I would say that if you have an older server as a TFTP server (i.e. dedicated to TFTP, not running the actual CallManager service) then I wouldn't worry about much else other than whether or not it's on the supported hardware list. As the TFTP service is throttled back with regard to the number of concurrent requests anyway, you should still get better performance by dedicating a box to this and upping the concurrent requests service parameter (which lots of people forget to do).

Similarly if you have a dedicated Publisher, (not running the CallManager service) I'd be less concerned about fairly minor differences.

In the end it simplifies things if you have consistent hardware, as it's one less thing to take into consideration when planning your resilience and so on; and in future when you do the next upgrade and a few of your servers drop off the supported list and you can't upgrade it...

Aaron

Aaron has covered one of the key points I was going to bring up, so no need to repeat. But to answer your question about the SRND statement concerning

" Cisco recommends that you use the same hardware platform for the TFTP severs as used for the call processing subcribers, page 8-13"

They are referring to running different classes of machines. Meaning, running a solution that has a mix of 7845s/7835s/7825s/etc. is a problem. I would never recommend doing that. As previously pointed out, even if a machine is just providing a function that has low overall CPU utilization (like TFTP), that machine is still a member of the cluster. So, it has to keep up with the size of the database and the ICCS traffic. If you find yourself in this boat you should try to get the cluster to a uniform platform. If that is simply not possible, then take the smallest platform and design all max capacities to that platform. Keeping in mind that (a) certain platforms do not support clusters with more than two nodes and (b) not all platforms are high-availability platforms.

Now, if you have a mix of IBMs and HPs (but all on the same platform, e.g. mcs7835) then this is not a problem in my opinion (and it appears you have agreement from everyone on this point), as long as you stick to some of the basic guidelines Aaron laid out (e.g. call processing servers should be the same hardware spec). Basically, even if they both are MCS7835 machines it is guaranteed one vendor's 7835 will outperform the other.

HTH.

Regards,

Bill