[Hawkular-dev] Response time ?

Thomas Heute theute at redhat.com
Fri Feb 20 05:53:17 EST 2015



On 02/20/2015 10:19 AM, Thomas Segismont wrote:
> Percentiles and standard deviation are needed in general.
>
> Now on a *default* dashboard (this is what you're asking about,
> correct?) I personally like to see avg AND 90 percentile.

Yes, really as overview information. Somehow indicator of the health of 
the monitored resource. Ideally something we can link to a color 
(Green->Red) for quick summary.

Thomas

>
>
> Le 19/02/2015 15:52, John Sanda a écrit :
>> I like percentiles. Here is a good article that explains why percentiles
>> are effective and why they should be used for baselining.
>>
>> http://apmblog.dynatrace.com/2012/11/14/why-averages-suck-and-percentiles-are-great/
>>
>>> On Feb 19, 2015, at 9:20 AM, Thomas Heute <theute at redhat.com
>>> <mailto:theute at redhat.com>> wrote:
>>>
>>>
>>> What is a useful representation of a response time over a period of time ?
>>>
>>>      - Average of all the samples ? *2.22s*
>>>      - 90/95/99 percentile ? 99% of your requests took less than *2.45s*
>>>      - Percentage of valid samples ? *94%* of your requests took less
>>> than your 3s acceptable time.
>>>      - Apdex ? http://en.wikipedia.org/wiki/Apdex
>>>
>>> Anything else ? (This is for a quick overview of the health of a system).
>>>
>>> Thomas
>>>
>>> _______________________________________________
>>> hawkular-dev mailing list
>>> hawkular-dev at lists.jboss.org <mailto:hawkular-dev at lists.jboss.org>
>>> https://lists.jboss.org/mailman/listinfo/hawkular-dev
>>
>>
>>
>> _______________________________________________
>> hawkular-dev mailing list
>> hawkular-dev at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/hawkular-dev
>>
>
> _______________________________________________
> hawkular-dev mailing list
> hawkular-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hawkular-dev
>


More information about the hawkular-dev mailing list