Evaluating server hardware: a sign of the times

re: Evaluating server hardware: a sign of the times

Thomas Kejser — Tue, 03 Jan 2012 11:31:38 GMT

Hi Linchi

Interesting benchmark and data here. I would really like to see the dump of waits/spins/latches (no filters of "ignore waits" please) for 1 minute of this run. Could you mail my address with it please?

Also, curious to know if you see large kernel times on this system?

Cheers and happy new year

Thomas

re: Evaluating server hardware: a sign of the times

GrumpyOldDBA — Tue, 03 Jan 2012 13:33:02 GMT

One of the key factors in comparing servers is the memory speed , for in memory iops this will be the limiting factor - on the basis there's always one part of the system which is a bottle neck.

Did you factor this in at all - just interested.

re: Evaluating server hardware: a sign of the times

Linchi Shea — Tue, 03 Jan 2012 14:51:05 GMT

Thomas;

I'll email you the stats when I get a chance.

GrumpyOldDBA;

Memoy speed is definitely a key factor, actually a major factor in determining the server performance, and the benchmark is shown to be sensitive to that.

re: Evaluating server hardware: a sign of the times

Daniel Bowers — Tue, 03 Jan 2012 19:06:08 GMT

Very interesting, thanks for posting this.

What were the "transactions per second" seen in your four E7-4870 scenarios? I'm curious to know whether that correlated with the "% Total CPU Usage" you measured.

Do you typically modify power management settings (in either BIOS or Windows) when running your tool?

re: Evaluating server hardware: a sign of the times

Linchi Shea — Tue, 03 Jan 2012 19:37:28 GMT

Thanks Daniel! I knew someone would ask the question about the throughput. I hope to have that coveredina next post as I can't really do justice in a comment. But suffice it to say that E7-4870 is impressive in the tests I have run.

The server is set for performance not power savings. I don't change the setting in any of the tests.

Performance impact: hyperthreading for OLTP queries

Linchi Shea — Fri, 06 Jan 2012 04:44:49 GMT

My previous post focuses on the performance impact of enabling hyperthreading (HT) on a machine with

Performance impact: hyperthreading on Intel Westmere-EP processors (X5690)

Linchi Shea — Sun, 22 Jan 2012 06:41:26 GMT

Recently, I have been looking into the performance impact of enabling hyperthreading on various platforms