Trying to use a tool like iostat against a shared, network provided block device to figure out what your level of service your database is getting from the filesystem below it is an exercise in frustration that will get you nowhere.
This may be true under Solaris. Since 2.5 Linux has had /proc/diskstats and an iostat that shows the average i/o request latency (await) for a disk, network or otherwise. For EBS it's 40ms or less on a good day. On a bad day it's 500ms or more if your i/o requests get serviced at all.
Comments
Trying to use a tool like iostat against a shared, network provided block device to figure out what your level of service your database is getting from the filesystem below it is an exercise in frustration that will get you nowhere.
This may be true under Solaris. Since 2.5 Linux has had /proc/diskstats and an iostat that shows the average i/o request latency (await) for a disk, network or otherwise. For EBS it's 40ms or less on a good day. On a bad day it's 500ms or more if your i/o requests get serviced at all.