When 90th% response time is higher than max

34 views
Skip to first unread message

JJV

unread,
Mar 5, 2012, 8:47:15 PM3/5/12
to faban...@googlegroups.com
I'm curious why sometimes faban runs show a 90th% response time which
is higher than the max response time?
For example:

Avg=0.389
Max=0.558
SD=0.065
90th%=4.500

Thanks!

--
Jyri J. Virkki - jy...@virkki.com

Sucharitakul, Akara

unread,
Mar 6, 2012, 1:03:59 AM3/6/12
to faban...@googlegroups.com
The max is a straight max but the percentiles are calculated from buckets.
The bucket granularity is calculated your percentile limits. This can
deviate the numbers.

-Akara

JJV

unread,
Mar 8, 2012, 2:06:24 AM3/8/12
to faban...@googlegroups.com
On Mon, Mar 5, 2012 at 10:03 PM, Sucharitakul, Akara
<asucha...@ebay.com> wrote:
> The max is a straight max but the percentiles are calculated from buckets.
> The bucket granularity is calculated your percentile limits. This can
> deviate the numbers.

Thanks! So sounds like best practice is to set the limit slightly
higher than the expected max to get the best bucket granularity. I had
the limit set way off (cut & paste from an earlier experiment which
had very different response times).

Sucharitakul, Akara

unread,
Mar 8, 2012, 3:10:05 AM3/8/12
to faban...@googlegroups.com
Yes, please set it properly so it is really your SLA. Thanks.

-Akara

Reply all
Reply to author
Forward
0 new messages