Iozone: Performance Test of File I/O Version $Revision: 3.308 $ Compiled for 64 bit mode. Build: linux O_DIRECT feature enabled Auto Mode File size set to 262144 KB Record Size 4 KB Record Size 64 KB Record Size 512 KB Command line used: iozone -I -a -s 256M -r 4k -r 64k -r 512K -i 0 -i 1 -i 2 Output is in Kbytes/sec Time Resolution = 0.000001 seconds. Processor cache size set to 1024 Kbytes. Processor cache line size set to 32 bytes. File stride size set to 17 * record size. random random KB reclen write rewrite read reread read write 262144 4 38211 42784 46048 45918 8684 37782 262144 64 69923 77192 115359 113808 91553 74527 262144 512 77909 60055 220997 221988 204858 77677
Through Google I found comparable benchmarks. This one was posted on the Ubuntu Forums. It should be noted that this is a 160 GB X25-M, the poster mentions that his one is a ‘G2’ which means that it’s a second generation one with the Postville code name like mine. Probably the greater amount of storage would have some benefit for performance, but I’m not sure. These numbers are taken from the benchmark without TRIM. If I understand correctly it doesn’t matter if you use an X25-M with the latest firmware which supports TRIM (like I do), because there is no support for it in Linux/Ubuntu yet and it looks like it won’t be in the next Ubuntu release either. With Google or in that topic you can find an explanation on how you could use a recent version of hdparm and some kind of trick to use TRIM, so that’s how that poster probably got his follow-up benchmarks with TRIM. I didn’t bother because I think I’d rather wait until the support for TRIM is mature enough for it to work out of the box. The poster used an HP Elitebook 8530p with an Intel Core 2 Duo T9400 CPU and 4GB DDR2-800 RAM.
random random KB reclen write rewrite read reread read write 262144 4 55854 61601 77975 77408 18740 37199 262144 64 102575 87223 200613 201029 141870 70205 262144 512 110951 93840 244588 242498 233184 95013
I found a second post with benchmarks on another forum, but unfortunately no more than that. In the specific post I just linked it’s not mentioned, but in an earlier post in the same topic the poster gives the model number of his SSD, INTEL SSDSA2M080G2GC, which means he has the same model I have. He posted his benchmarks at 5 December mentioning that it they were made with the most recent firmware. If I recall correctly, that’s still the latest firmware at this moment. So he’s using the same firmware as I am, the first firmware to include support for TRIM. Not sure what system was exactly used for the benchmark, but the poster mentions it’s a notebook. I’ve asked him and I’m waiting for a response.
random random KB reclen write rewrite read reread read write 262144 4 39360 46785 53897 51421 11412 39529 262144 64 71098 54363 130520 129911 98805 74485 262144 512 81657 78925 207837 210842 218651 81242
So which conclusions can be drawn from this? No definitive. I should also take into account I’m using the EXT4 filesystem on my X25-M in combination with an alpha version of Kubuntu 10.04, which uses the 2.6.32 kernel. Benchmarks done by Phoronix show that with this and other recent kernels EXT4 suffers from performance regressions. The numbers presented by the benchmarks done by the poster on the Ubuntu Forums leave my X25-M in the dust, but comparing to the last benchmarks doesn’t give such a dramatic difference. The greatest difference can be found in the benchmark with the 4KB blocks (first row). If anyone has a better interpretation of these benchmarks and the context to offer, please comment.
Edit 12 February 2010: the following results were achieved with the ext3 file system, using the noatime option. Contrary to my expectations it’s not better, but sucks more.
random random KB reclen write rewrite read reread read write 262144 4 45452 42928 43576 43587 9120 40782 262144 64 59020 73224 109954 108622 90028 74706 262144 512 81555 82546 172479 171934 174380 82114