Comments on: iostat and disk utilization monitoring nirvana

By: Travis Bear

Travis Bear — Tue, 27 Apr 2010 20:34:09 +0000

Excellent explanation. Just what I needed. Thank you!

-Travis

By: Thiru

Thiru — Tue, 29 Sep 2009 14:12:52 +0000

Great. . . Gud keep it up. . .

By: Abid

Abid — Thu, 16 Jul 2009 11:10:11 +0000

Good article though I’m confused whether or not it applies to an issue at my hand.

To put it simple, I’ve a java application wherein I spawn some 200 threads and each thread has to read/write its own file inside one common directory say “temp”. However I’ve a requirement that only one thread (and in terms of latest linux kernel I may say each process) do read/write operation on “temp” at a time. So, I streamline the operation by writing a code that can be represented by following pseudo code:

lock(“temp”){

do read/write

}

So, now if 200 threads concurrently try to execute this code they won’t be able to do so; one thread will execute this piece of code at a time and others will get blocked.

So, I was just wondering if thread-1 starts now and say after 3 seconds thread-200 gets to acquire the lock and perform read/write operation, will this waiting time of 3 seconds affect “await” of iostat?

By: Atsushi

Atsushi — Tue, 31 Mar 2009 07:30:13 +0000

Very clear description of %util of a disk.
It is totally what I’ve looked for.
Thank you.

By: tinkertim

tinkertim — Thu, 12 Mar 2009 14:15:29 +0000

Have you checked out dm-ioband? Not only are the metrics easier to obtain, but great QoS can be achieved.

By: N

Mon, 09 Mar 2009 12:36:15 +0000

Very few web hosting companies seem to do a detailed analysis. For shared hosting, you can see a lot of variation in response time with even the larger web hosting companies.