David Thompson wrote:
>
> We throw caution to the wind and run write-cache-enabled without a BBU, and
> still see a significant drop in performance compared to raid-1.
Well yeah, how on earth can you compare the performance of a raid-5
array with a raid-1 array? raid-1 write performance is the same as a
single disk (within reason).
> Let me know what you find.
The BBU showed up today and managed to charge prior to the arkeia backup
window so I could turn on write cache. At a glance statistics comparing
Saturday night to now:
sat:
CPU %user %nice %system %iowait %idle
08:10:05 PM all 4.88 0.91 3.16 32.63 58.41
08:20:10 PM all 6.69 0.43 4.25 44.49 44.15
08:30:09 PM all 9.95 0.64 5.74 50.90 32.76
now:
CPU %user %nice %system %iowait %idle
08:10:01 PM all 1.97 3.64 2.93 9.19 82.27
08:20:01 PM all 2.62 5.42 4.30 12.79 74.86
08:30:02 PM all 3.49 3.91 4.82 16.00 71.78
I'd say that was a rather significant increase; it's in no way a formal
test, but if I see this type of result after a few more nights then I'd
say the $100 for the BBU to enable write cache safely is money well spent...
-te
--
Troy Engel | Systems Engineer
Fluid Inc. | http://www.fluid.com
