Total Transactions per second (or Hits/sec or Throughput) Graph anomaly

[GMSRao]

Hi,

We have configured a load test to run at a certain TPS during steady
state. (after timing each of the actions/Txs with think time etc). At
steady state we do see the targeted TTPS...but there is a weird
behavior we see with TTPS or Hits/second or Throughput graphs when the
results are opened in 'LR analysis' when it comes to certain
granularity. For example, I see a reasonably steady graph for TTPS
(at targeted level ..per say 12 TPS) when filtered for steady
state...but the same results when I change granularity to say '20'
seconds or multiples of it, I see spikes in TTPS graph throughout (for
every 3 minutes)...like a zig-zag graph..at every 3 mins. Has anyone
come across such a case and if so, any suggestion to overcome would be
helpful..

[Performance Testing]

Hi,

As you reduce the value in the granularity option in LR analysis, the spikes are bound to increase.

The interval at which each sample is displayed is the granularity of the graph. A similar but not exact analogy exists with scale in maps. The larger the scale, you will see smother edges in the graph and if we reduce the scale, more and more rough edges on the map can be seen.

To put it differently, if you have granularity of 5 mins, LR will get a mean value of all the samples collected in that 5 min. In an hour scenario, you can only see 12 data points. In other words, you have 12 data points which are a mean of the samples collected during each interval of 5 min. However, if the granularity is reduced to 1 min,  total of 60 data points can be seen on the graph. This translates to 60 data points which are the mean of samples collected during each interval of 1 min. In almost all cases, the variation in 5 min sets of mean will always less than the variation in 1 min sets of mean. This explains the spikes you are seeing.

This is the expected behavior of any tool that displays graph. Let me know if I need to make it more clearer.

Note: The rate at which samples are collected is called the sampling rate, while rate at which samples are displayed is the granularity. So in any case you can never have granularity less than the sampling rate.

Cheers,

Perf Trainer

--
You received this message because you are subscribed to the Google "LoadRunner" group.
To post to this group, send email to LR-Loa...@googlegroups.com
To unsubscribe from this group, send email to
LR-LoadRunne...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/LR-LoadRunner?hl=en

[GMSRao]

Thanks for the reply explaining on granularity and sample rate.

The weird thing is when i set the granularity to 18 seconds (the
minimum i am able to set for this particular run) i do not see the
spikes, its smooth. But when i set it to 20 seconds or multiples of
it i see spikes (double the max in 18 seconds graph).

On May 22, 8:25 am, Performance Testing

[GMSRao]

When i look at the graphs mentioned using 'summary data' i dont see
any of the 3 minute interval spikes on these graphs (TTPS, TPS, Hits/
sec, Throughput). But when i reload the data with 'complete data'
then again these spikes reappear (could this be some issue with 'data
aggregation' algorithms used within LR analyis)

> > >http://groups.google.com/group/LR-LoadRunner?hl=en- Hide quoted text -
>
> - Show quoted text -

[Performance Testing]

aaha....

Complete data is generated after the raw data gets processed in LR Analysis engine. So graph showing the complete data is considered to be more reliable and ideally you should only look at complete data while you are doing analysis.

 

Coming to the TTPS thing. in a practical scenario, you cannot expect a smooth graph even if everything else is perfect. There is bound to be such spikes during the entire scenario. If you can send me analysis file and the scneario details, I can explain you better.

 

 

Cheers,

Perf Trainer

--
Cheers,

Perf Trainer