If you are unable to create a new account, please email support@bspsoftware.com

 

News:

MetaManager - Administrative Tools for IBM Cognos
Pricing starting at $2,100
Download Now    Learn More

Main Menu

Server Sizing on a PVU Based Licensing Model

Started by locus, 26 Jun 2015 06:46:47 AM

Previous topic - Next topic

locus

Hi Folks,

We have a TM1 Server that's been around since the days of server licensing for the application.
Evolution of the product capability and marketing has driven changes - at least where we are - in the costing model, now based on PVU points.
These are calculated from the physical machine specs and how many cores we will use for our partition - Running VMware.

Whether we will be able to transfer or just flat out pay for cores is yet to de determined.
The current spec of 8-Cores (With Plenty of memory) is definitely 'Bloat', looking at the utilisation stats for the last 3 months.
It is still a very lightly used system - this will be changing quite soon -  and I am of the opinion that 4-Cores is plenty.
To date peak CPU utilisation is 20%.

We are however planning to run 2-Cores - and the same MEM as prod - on our Non Production servers.

Prior to any applied discounts , we are looking at around $100k per core, so this is a decision I want to get right.
No bloat but no under spec. If we need to push cores later, 2+2 will cost a lot more than 4, because of volume based discounting.   ;)

Do any of you have hands-on experience running 2-Cores in production with a complex, fairly large model and an ever-growing use base?

Am I right to think it's an under-spec?

Cheers,
Locus

bdbits

I think it is difficult to say for certain, as a lot depends on how many concurrent users you have, and report servers needed and defined, and the nature of the data sources and reports themselves.

If you have server utilization statistics available (sounds like you do), I would take a hard look at what they are telling you. I do not have privileges to look at our prod server definitions, but I believe we are running 4 CPUs on each of our two VMware-based production nodes. The server admin has told me many times that most of the time, we are not at all CPU-bound on prod. On the other hand, our dev and test servers are single VMs with I think 2 cores each, but run considerably slower under any kind of load. So my personal feeling is that unless you have very low concurrent usage, 2 cores might be too light. But it is just my instinct and is not based on experience other than what I have just said.

Your IBM rep should be able to put you in contact with IBM technical services to assist you in sizing for your particular circumstances. These are not sales people, and I have found they generally will give you advice without trying to up-sell you. We never get charged for their advice - well, not directly, I am sure this is built into our annual software renewals.  ;D