1 Reply Latest reply on Nov 26, 2019 11:05 AM by Seth Paskin

    How much machine learning are we getting with True sight?

    Jamil Hussain
      Share This:

      I have started a masters research degree in machine learning and analytics - given the little i now know!!

       

      I think the BMC product Truesight is an excellent first step at probable root cause however there is so much more we can do I know we get machine learning in BMC truesight, e.g.probable root cause analysis, my question having done BMC products most of my working life is how much more data can we and should we be doing.

       

      Remember machine learning is about looking at data and then making some hypothesis on that data applying regression and fitting models and then doing something with that data, now given that we collect a lot of data with patrolagents (cycles collect data every 2 minutes - UNIX KM collects over 100 parameters/ORACLE KM collects over 200 - parameters) given we have all this data collected, my point is:-

       

      how much statistical modelling are we doing?

      how much machine learning are we doing?

      Is there any plans to improve this, i say this because on my research program there is so much we can do, with this data, lets take for example the unix KM now if we are collecting the 1 minute load average, we can do a correlation analysis with this value and the CPU or memory or disk load to see what the correlation is. this would be straight forward to do. similar we could do this with many other parameters and collections.

      is there any plan to provide some sort of big data capability for tall this data that we have been historically collecting?

      Are we as customer being handicapped at the license costs at every KM? approach, this maybe a question for not this discussion but i am raising anyway, if you want to do machine learning and carry out intelligent decisions etc, then you need as much data as you can or as many (samples from that data), if you are not collecting those for license issues etc then you dont have that data to make those decisions. ( the licensing model needs to be easy to work with) is the point.

       

       

      questions/thought/answers

       

      Finally Well done for such a good start with Probable root cause, (and i think it has the sweet somewhat in users do not need to be experts to use machine learning!!