At around 60 to 65% Little’s law kicks in and you’ll start to see a huge increase in response times. So when you see disk utilization of almost that value you should be hearing alarm bells to warn you of upcoming performance degradation. Remember that a disk can have 180 IOps (15000 RPM), but this will be 100% utilization and response times of individual IOs can be very large. I always try to stay at 2/3 of the max, everything above that can be noticable on critical servers as response times will go up exponentially.
I would certainly move one of the luns away off on too another set of disks, one count for improved performance and another it is not best practice to have Luns with high throughput on the vault drives.
RRR
4 Operator
•
5.7K Posts
0
April 14th, 2011 07:00
At around 60 to 65% Little’s law kicks in and you’ll start to see a huge increase in response times. So when you see disk utilization of almost that value you should be hearing alarm bells to warn you of upcoming performance degradation. Remember that a disk can have 180 IOps (15000 RPM), but this will be 100% utilization and response times of individual IOs can be very large. I always try to stay at 2/3 of the max, everything above that can be noticable on critical servers as response times will go up exponentially.
natsidan
1 Rookie
•
34 Posts
0
April 14th, 2011 13:00
I would certainly move one of the luns away off on too another set of disks, one count for improved performance and another it is not best practice to have Luns with high throughput on the vault drives.