5 Replies Latest reply on Oct 22, 2012 8:39 AM by dsbalcau

    SMI-S Configuration and performance data intervals

    Deltona

      Hi guys,

      I have a question regarding the statistics that STM gathers via SMI-S providers. I know that there’s the default 15 minute intervals to collect data and that one could tweak the systems in order to gather statistics more frequently like say 5-6 minutes instead.

       

      Now, what I don’t get is what period of time is being listed in the statistics gathered every 15 minutes? Are these values from a specific point in time, an average, a peak, from the collection time or other?

       

      In other words, do the perf stats gathered every 15 minutes reflect on statistical data gathered every minute, 15 minutes or something elsea altogether?

       

      Looking at the charts I can only see 15 minute intervals and I am not sure whether the values presented at each 15 minute interval are averages, peaks or the actual performance value at that specific collection time…

       

      I don’t really want to base thresholds on something that isn’t well documented. In reality, a threshold could be set for something that may not be reached if that value is an average and not a constant.

       

      Last but not least, in an HP EVA setup, who or what decides what data and at which intervals is being made ready to the SMI-S provider? Is it the EVA?

       

      Regards,

      Deltona

          • Re: SMI-S Configuration and performance data intervals
            bmrad

            What we get from SMI-S provider is cumulative counters.  When 15 minutes frequency is used, we get two samples 15 minutes apart and take the difference.  Now if you take any of the xxx/sec values, these will be an average on the 15 minute interval. It will not catch the peaks that happened within those 15 minutes.

            SMI-S provider reports it as counters, not as point in time values or graphs for a specific interval.  Storage Manager communicates to the EVA provider and it in turn talks to the array. 
            It is pull operation from the EVA Provider side.  Provider decides how often it goes and gets the data.  It can be as frequent as on every request from a client. Or, it can be based on a frequency if there is provider level caching.

              • Re: SMI-S Configuration and performance data intervals
                dancoh-betsson

                Care to simplify your answer? Can STM only get performance each 15min? Or does it only collect every 15minutes but the data it collects are for the complete 15min? Cause in command_view you can get data in realtime or at least each minute. Why not in STM? You are getting an ok overview by collecting each 15min but you don't see the peaks.

                  • Re: SMI-S Configuration and performance data intervals
                    dsbalcau

                    Hi dancoh-betsson,

                     

                    I'll try to clarify some of the points:

                     

                    Can STM only get performance each 15min?


                    This is limited by the SMI-S provider. We can query the data more often from the STM side, but the Provider's are designed to not affect array performance and only query new data from the array every 10-15 minutes. Therefore, even if STM asks for data more often, we are not guaranteed to get "real-time" data.


                    Cause in command_view you can get data in realtime or at least each minute. Why not in STM?


                    Command_view is not using the SMI-S provider. STM could potentially get more real-time updates from the array if we parsed command line output from the array, but this is a non-standard practice that requires custom code to be written for each array we support. It also is not a robust way to write software, as it can break if the array vendor's decide to change how they output data.