Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

This question already has an answer here:

I have a Linux Mysql server cluster where master and slaves are overloaded by a combination of read/write i/o and by SELECT queries load.

We purchased FusionIO cards to replace the hard drives. My question is whether to upgrade the servers themselves, or just replace the hard drives by FusionIO cards in the existing servers.

The existing servers have 266 MHz memory (Xeon E5345 - a bit outdated by now), while we can purchase servers with at least 1333 MHz RAM. We would rather not spend money on new servers, since FusionIO cards are already very costly.

We tested a FusionIO card in a brand new 1333MHz RAM server, and it gave us 4-5x speedup: but we do not know the contribution of RAM speed vs FusionIO.

What are the best practices on Linux to examine whether the RAM speed is a real bottleneck or not?

share|improve this question

marked as duplicate by Michael Hampton, HopelessN00b, EEAA, MDMarra, mdpc Apr 1 '13 at 3:29

This question was marked as an exact duplicate of an existing question.

You bought the FusionIO cards before you had any idea whether they would help or not? – Michael Hampton Apr 1 '13 at 2:20
@Michael Hampton - how did you make that assumption? I bought two test FusionIO cards and put them in brand new servers with 1333MHz RAM, and that gave us 4x speedup. What we do not know is what was the role of FusionIO versus RAM speedup. Thus the question. – mercador Apr 1 '13 at 2:32
Because you never mentioned anything about doing any tests. – Michael Hampton Apr 1 '13 at 2:33

New servers > FusionIO on old servers...

Memory bus speed is probably not your bottleneck. The Intel 5300-series CPUs are four generations behind the currently-available CPUs.

Architectural differences as well a two MAJOR shifts in CPU technology (going from Harpertown to Nehalem and later, Westmere to Sandy Bridge) would account for most of the performance differences you may have experienced.

share|improve this answer

Not the answer you're looking for? Browse other questions tagged or ask your own question.