Frankly, at this point I'm more concerned about end-user needs than specific features like the ones you mention.
Input data could be XML or CSV -- but we could accomodate and have JMeter produce something else, or have a transformation step.
As I said, it's mostly working with multiple sets of timed samples. That is: JMeter runs a load test script against (say) a web app, and produces one timed sample for each request, containing essentially the sampling time and request/response time interval (latency). We could view the set of samples we've obtained for each page as a separate sample set.
Without much thought:
- Creating a sample set from the union of several other sets. - Displaying the samples in a latency vs. time graph, possibly combining multiple sample sets in the same graph. - Grouping samples in a set by time intervals (minutes or hours) and displaying averages and standard deviations and/or certain percentiles, max and min. Again, possibly combining multiple sample sets in the same graph. Which kind of graph(s) is not important, as far as they are sufficiently readable. - With the same grouping as above, compute the throughput rate (samples per time interval) and display it in that same graph. - With the same data as above, show a latency vs. throughput graph (instead of the latency & throughput vs. time I described before). A fitting curve would be nice, but it's not a must.
Zooming and scrolling could be necessary in the time axis, since tests can span very long periods of time and, sometimes, the period may be interesting.
Note that we already have part of this (rather ugly, but essentially working) in JMeter. We could continue to develop it that way. I was thinking about using something more generic to allow users to play with their data and look at it from different angles. Also because we may be adding other measurements in the near future (CPU load on the different servers, memory usage, Garbage Collection events, network load, etc... -- plus others we already have such as the response size) and pre-defining all the possibly interesting graphs and views is nonsensical, IMO.
-- Salut,
Jordi.
J.Pietschmann wrote:
Jordi Salvat i Alabart wrote:
I was actually looking for something readily available we could "partner" with -- rather than develop it ourselves.
Hm, could you give a more detailed list of features you expect? E.g.: - input data structure and format (is it a file?) - what statistics? fitting curves? clustering? calculating medians, standard derivations, quantiles, rankings? test for abnormal data? - what nice graphs? lines, bars, different markers, candles, 3D? - just plot or do user interaction: zoom in etc?
J.Pietschmann
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
