|
Disk I/O too fast?? (SATA-2)
Hya,
While I am tuning up my Penguin (Debian, squeeze), I am confused.
Main job with this penguin is processing huge data, so I guess so so CPU intensive and disk I/O. "top" reading agrees, 40-80% user CPU, 5-10% iowait.
Next, I wanted to check whether disk i/o was a bottleneck, in this case putting data on SSD may improve performance. "iostat" reading says that ca. 180000 blocks read/sec. Now, I am lost. "tune2fs -l" says that 1 block = 4096. (mkfs says that unit is byte, not bit).
1.6E4 block_read/sec $times$ 4096 bytes/block $times$ 8 bit/byte $nearly equals$ 5.9E9 bit/sec. This exceeds SATA-2 spec (which is 300MByte/sec = 2.4E9 bit/sec), of course disk transfer rate in catalog, (1406 Mbit/sec).
What is wrong? Answer is that the block iostat refers is 512 byte as man page says!
So transfer speed is 5.9E9 bit/sec $dev$ 8 $equals$ 7.4E8 bit/sec.
If anybody else run into same question, read this thread!
Happy Penguins!
|
