Startup times with new architecture

Post Reply
brandon.barker
Posts: 5
Joined: Mon Jan 11, 2010 4:00 pm
OLAP Product: TM1
Version: FP2 hotfix 32
Excel Version: 2007

Startup times with new architecture

Post by brandon.barker »

I was wondering if anyone has had a chance to put TM1 on a newer box that has maybe the new Intel Nahalem processors with DDR3 memory or similar. Did you see a startup time decrease? I know TM1 is loaded into memory and was wondering if hardware would be that big of a factor.

Thanks!
User avatar
Martin Ryan
Site Admin
Posts: 1989
Joined: Sat May 10, 2008 9:08 am
OLAP Product: TM1
Version: 10.1
Excel Version: 2010
Location: Wellington, New Zealand
Contact:

Re: Startup times with new architecture

Post by Martin Ryan »

While the information is pre-loaded into RAM during startup, the calcs and feeders must be worked out by the processors before they can be put in RAM. As a result you will notice performance changes depending on both your CPU and RAM setups.

Cheers,
Martin
Please do not send technical questions via private message or email. Post them in the forum where you'll probably get a faster reply, and everyone can benefit from the answers.
Jodi Ryan Family Lawyer
image2x
Posts: 125
Joined: Tue Jun 02, 2009 7:05 pm
OLAP Product: TM1, PAX, PAW, SPSS
Version: 2.0.916.10 on RHEL
Excel Version: 2016
Location: Minneapolis, MN

Re: Startup times with new architecture

Post by image2x »

Well here's my recent experience...

Our dev and prod machines are both both dual X5570's (Nahalem generation). We were running 128GB in both until I saw that the memory rate was only at 800Mhz when I knew the processor and board were capable of 1333Mhz. Turns out that on DL380 G6 Proliants, if you want to run at 1333Mhz, you can't have the third channel populated.

So, eagarly hoping to lower our 55 minute startup (and 30-40 minute, lockout security refreshes on security property changes), I dropped the ram to 96GB (8x12) and bumped the speed in the bios up to 1333Mhz. The result was disappointing (at least for server startup)... there was zero change -- still at 55 minutes. Our startup is limited to only 1 thread so perhaps things might have been different with more threads.

That said, things (websheets in particular) do seem a little snappier but I really can't point to the memory speed increase as a definitive improvement for end users or for tm1 processes.

-- John
Post Reply