topic Re: TPM Ratings in Operating System - HP-UX

TPM Ratings

Darren Murray_1 — Fri, 30 May 2003 23:54:14 GMT

Hi ,

A proposal was put in front of me of running a 64way superdome and carving it up into 4 hard partitions, with one of those hard partitions having 2 vpars. The first Vpar will have 8cpus and 16Gb RAM. The second Vpar will have 4cpus and 8Gb RAM.

What is proposed to run on VPAR1 currently runs on 14 x L1000's with 2 x 440Mhz Cpus.

What is proposed to run on VPAR2 currently runs on a N4000 with 6 x 550Mhz

My TPM rating spreadsheet tells me that what is currently running has quite a lot more CPU power.

Can you guys please give me your opinions.

Much appreciated.

Cheers
Darren

Re: TPM Ratings

Pete Randall — Sat, 31 May 2003 01:55:09 GMT

Nope!

But, I do have contacts who know how to pronounce Superdome. I will query and respond if I can get any input. Interesting!

Pete

Re: TPM Ratings

Jeff Schussele — Sat, 31 May 2003 02:09:13 GMT

Explain how the first app runs *across* 14 systems?
Also what was the RAM count on those L1000s?
I believe the second scenrio is safe. What happens with the OTHER 3 cells. Why not one app one cell? npars are where the SD lives - vpars are inherently *more* overhead & you lose the benefit of independence - i.e. one down all down.

Rgds,
Jeff

Re: TPM Ratings

Darren Murray_1 — Sat, 31 May 2003 02:26:32 GMT

There is 2.5GB of Ram in each of the L1000's

At the moment the L1000's form 5 MC/Service Guard Clusters

4 x 3 node clusters and 1 x 2 node clusters.

All servers are active.

They all run a number (24) of separate oracle databases.

There will be other applications running on some other partions but I am only concerned about the these 2 VPARS. Im assuming that I can discount their affect.

Im new to the Superdome thought so

Cheers

Re: TPM Ratings

Michael Steele_2 — Sat, 31 May 2003 02:58:16 GMT

Power, I/O backplane, firmware upgrade and I also believe the system clock are all considered superdome SPOF. Here is a good thread about it:

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xac475c7609e9d61190050090279cd0f9,00.html

My understanding is Superdome is a more natural replacement for the V Class server and is therefore real gorilla for big kernels and systems with lots and lots of processes. So Superdome compliments V Class environments. (* I always thought partitioning kind of defeated this. *)

Also, memory cells and CPU's have their own oddities. Like, add memory to one cell requires that you add memory to all the cells. Here is good link:

http://docs.hp.com/hpux/onlinedocs/os/11i/superdome.pdf

I've also attached a Superdome Partitioning PDF.