iostat tps, what's alarming?

[hfaix]

I'm looking at days and days worth of I/O stat data trying to determine if I have an I/O bottleneck. Specifically, I've been asked to see if I have a disk hotspot.

In analysis I see two hdisks (out of 15) that have tps on occasion between 600 and 800. Is this bad?

I've tried using filemon, but running a trace all day isn't realistic...so I've opted to use iostat results.

 

[khalidaaa]

Hi hfaix,

I always look at the %tm_act to indicate whether the disk is overutilized!

% tm_act Indicates the percentage of time the physical disk was active
(bandwidth utilization for the drive).

If only two disks out of 15 are busy then why don't you try to spread the data over the 15 disks? This will give you higher throughput!

Regards,
Khalid

 

[hfaix]

Yep that's what I'm going to do. I was trying to avoid it if possible. I found out that 800 tps (IOPS) is pretty much all my disk subsystem can do.

thanks