For Day 27 of this series, I am going to talk a little bit about disk performance and one easy tool you can use to compare the performance of different types of disks and disk arrays.
CrystalDiskMark, available from Crystal Dew World is a widely used I/O component benchmark. You can select the number of test runs, desired file size, and which logical drive you are testing and it allows you to measure:
Sequential read and write performance in megabytes/second
Random read and write performance for a 512K block size
Random read and write performance for a 4K block size
Random read and write performance for a 4K block size with a queue depth of 32
There are many other ways such as SQLIO that will do a much more thorough job of benchmarking your disk subsystem, but they are a little more difficult to work with. Using CrystalDiskMark should be a supplement to other disk benchmarking that you do.
I have captured test results for a pretty decent consumer grade 6Gbps MLC SSD (the 128GB Crucial RealSSD C300) in Figure 1, compared to a couple of fast, 7200rpm SATA drives in Figures 2 and 3.
Both sequential and random read write performance are much better with the SSD drive. One thing to keep in mind is that most consumer SSD drives perform better in their larger capacities (so a 256GB drive will have better performance than a 128GB drive of the same model line). You also want to make sure that you use a 6Gbps SATA III port for a 6Gbps SATA III SSD, or else you won’t get get all of the performance the drive is capable of.
Figure 1: 128GB Crucial RealSSD C300 SATA III SSD
Figure 2: Western Digital Black WD1002FAEX SATA III hard drive
Figure 3: Western Digital Black WD1001FALS SATA II hard drive
Figure 4: CrystalDiskMark Scores for two 15K SAS drives in RAID1
Figure 5: CrystalDiskMark Scores for six 15K SAS drives in RAID10
Figure 5 shows the results from six 146GB Seagate Cheetah 15K SAS drives in a RAID10 configuration. Notice the 531.7 MB/s sequential read and 414.6 MB/s sequential write score. This shows the positive effect of having six drives compared to having four drives (using the exact same drives and RAID level). Even with six 15K drives in RAID10, the 4K random read and write performance is pretty low.
A single, consumer-grade Multi-Level Cell (MLC) SSD drive will have much, much higher random I/O performance than this six drive RAID10 array. This is shown if you compare random read and random write numbers between the six drive RAID 10 array in Figure 5 and the single, consumer-grade Multi-Level Cell (MLC) SSD drive in Figure 1.