IOMeter v2006.07.27 – IOps

IOMeter v2006.07.27 


Iometer is an I/O subsystem measurement and characterization tool for single and clustered systems. It was originally developed by the Intel Corporation and announced at the Intel Developers Forum (IDF) on February 17, 1998 – since then it got wide spread within the industry.

Meanwhile Intel has discontinued to work on Iometer and it was given to the Open Source Development Lab (OSDL). In November 2001, a project was registered at SourceForge.net and an initial drop was provided. Since the relaunch in February 2003, the project is driven by an international group of individuals who are continuesly improving, porting and extend the product.

Kingston SSDNow V Series 40GB Desktop Bundle Review (with RAID!) - Storage 28

Kingston SSDNow V Series 40GB Desktop Bundle Review (with RAID!) - Storage 29

Kingston SSDNow V Series 40GB Desktop Bundle Review (with RAID!) - Storage 30

Kingston SSDNow V Series 40GB Desktop Bundle Review (with RAID!) - Storage 31

Now for the good stuff.  IOMeter tests are great for showing what these drives can do when pushed to the limit.  Throughout all tests, the IOPS performance of the single SSDNow 40GB came out to roughly half that of its 10-channel brothers.  Even with half of its channels hobbled, the Intel controller was still able to outperform the OCZ Vertex.  This is mostly attributed to the Indilinx controller having very limited support for command queuing.  The dip in the first data point during the file server test was data scatter caused by the short test passes required as to prevent fragmentation from impacting the overall test results.

Different things happened when we tested our RAID pair.  They excelled in the pure random reads of our web server test, even beating the X25-M G2 TRIM model.  This was because the G2 is saturating the single SATA channel at its plateau.  In the file server test, the duo almost perfectly matched the 80GB G1.  Results became weaker and more inconsistent as we entered the write-heavy database and workstation tests.  This is actually an issue I’ve noted with SSD’s RAIDed under an ICH.  Either the ICH or its driver seems to amplify what would normally be small delays encountered as each SSD services a given write request.  I’ve noted this amplification gets even worse as you add more SSD’s to the RAID.  I’m still investigating this (as I have been for months now), and continue to work with Intel to try to get to the bottom of it.  What I want you to take away from the above is that the lack of scaling for mixed read+write tests under a RAIDed pair of SSD’s is not necessarily the fault of the drives.

« PreviousNext »