Fusion-io ioXtreme PCI Express SSD Review
As we noted in a previous SSD round-up article, though IOMeter is clearly thought of as a well respected industry standard drive benchmark, we're not completely comfortable with it for testing SSDs. The fact of the matter is, though our actual results with IOMeter appear to scale properly, it is debatable whether or not certain access patterns, as they are presented to and measured on an SSD, actually provide a valid example of real world performance for the average end user. That said, we do think Iometer is a solid gauge for relative available bandwidth with a given storage solution.
We've also considered two schools of thought with respect to proper Iometer testing from an unpartitioned, clean drive test setup and also a full, formatted volume that is obviously more representative of how an end user would setup the product. However, the latter case also fills the drive up with a large test file and then reads and writes to that file, which is less than a real-world usage model as well.
And so, as a result of all the back-and-forth discussion we've had around the lab lately with respect to this benchmark, we've decided to show you all the angles Here's a sampling of our test runs with Iometer version 2006.07.27 with the ioXtreme and the various competitive solutions we tested both in unpartitioned/formatted states and also with a full NTFS partition.
* Note: Clean, unformatted drive performance is represented in these tests
The above test condition with Iometer is representative of our "Workstation" test pattern performed on blank SSDs with no partition. Thinks of this test as a "best case" scenario in terms of performance for the test pattern since higher bandwidth is available on a completely clean SSD with a fresh low-level format, though again this is hardly how an end user would run the product in a real usage model. We also can't test the ioXtreme in RAID 0 here since the product relies on Windows software RAID with a formatted NTFS partition. Regardless, you can see the ioXtreme is posting available throughput that is some 4X faster than its closest competitor, the OCZ Z-Drive and about 5X the speed of a single Intel Gen 2 SSD.
Next we have the same test, only this time with a formatted NTFS partition which means we can also offer RAID performance numbers with the ioXtreme Pro and ioXtreme in RAID 0. You'll have to forgive us that we don't have RAID arrays with Intel's SSDs in these tests due to time constraints but you can still get a sense of relative performance.
Partitioned and Formatted NTFS Volumes Tested
Here a single ioXtreme, in the same test but on a formatted NTFS volume, offers about 28K IOPS, which is significantly faster than any other single drive solution in the test, as you can see. However, versus the peak of 41K IOPS we saw on the unformatted, clean ioXtreme, this is a decidedly different picture, though just as impressive perhaps. As a frame of reference, the Intel SSD dropped off about 1/2 of its bandwidth, though the OCZ Z-Drive managed to maintain much of its performance and even scaled up as we added outstanding IO requests. What's interesting is how much of a hit the ioXtreme took, though still dominant performance-wise. With the subsequent performance runs over time, perhaps the garbage collection algorithm of the ioXtreme, needs more time to recover versus the OCZ product, which is made up of Vertex series SSDs (and the new garbage collection algorithm that OCZ worked on with Indilinx) along with an LSI RAID controller.
