I'm trying to performance tune the I/O of my web server, which is at 41.1% reads merged (If my math is correct), which seems a tad high to just be going along with the defaults. Will modifying read_ahead_kb affect the value of "reads merged" in diskstats? If not, what's a good way of tracking performance progress so I know when I've set read_ahead_kb too high?
Field 2 -- # of reads merged, field 6 -- # of writes merged
Reads and writes which are adjacent to each other may be merged for
efficiency. Thus two 4K reads may become one 8K read before it is
ultimately handed to the disk, and so it will be counted (and queued)
as only one I/O. This field lets you know how often this was done.
Merging reads is an efficiency optimization, and if it's happening often, that's a good thing -- your disk is doing fewer reads, fewer seeks, than it would have been otherwise. Too much seeking can badly impact the performance of a mechanical disk. SSD's not so much.
It's happening because your system is doing a lot of reads that are close together for some reason. There's not really any deeper meaning than that.
So is the "reads merged" inclusive or exclusive of reads that were "merged" because of read_ahead_kb? If reads merged isn't a good indicator of when I have read ahead set to the sweet spot then what would be the metric(s) to watch?