5 Replies Latest reply on Nov 7, 2013 8:41 AM by Gregor

    PI Point Statistics

    mikeloria

      I have 30K PI points and I have multiple people setting up points. I wish to be able to determine the footprint for each point (the rate of events per minute for each point). I want to find which ones have the highest rates so I can possibly set a compression that is logical.

       

      I am considering righting a C# program to do this using OLEDB but I am wondering if there is a tool or a canned way to already do this? Thank you very much for your help in this matter

        • Re: PI Point Statistics

          Have you looked at compression insight?

            • Re: PI Point Statistics
              mikeloria

              I have not looked at compression insight. Please provide more details. Is this a tool freely available? Is it an OSIsoft tool or a third party tool? Thank you very much for sharing.

                • Re: PI Point Statistics

                  Here you go Mike: partners.osisoft.com/solutions.aspx

                   

                  It is a 3rd party solution from Pattern Discovery.

                    • Re: PI Point Statistics
                      dlee

                      You can use the pidiag -archk <path to archive file> command line tool to see a summary of archive statistics. The archive will have to be unregistered in order to run the check.

                       

                      The command will show you something like this:

                       

                      Analyzing archive: C:\Program Files\PI\Queues\FLORY_2013-02-12_00-00-00.arc

                       

                      -------------------------------------------------------------------------------

                       

                      recno:      1 id:       1 indices:  1 records:   30 events:      4238 fr: 99.1%

                       

                      recno:      2 id:       2 indices:  1 records:   30 events:      4227 fr: 98.8%

                       

                      recno:      3 id:       3 indices:  3 records:  415 events:     62771 fr: 99.4%

                       

                      recno:      4 id:       4 indices:  8 records: 1005 events:    493550 fr: 99.4%

                       

                      recno:      5 id:       5 indices:  3 records:  294 events:     48617 fr: 99.7%

                       

                      recno:      9 id:       9 indices:  1 records:   37 events:      9180 fr: 99.4%

                       

                      recno:     10 id:      10 indices:  2 records:  129 events:     31983 fr: 99.6%

                       

                      recno:     11 id:      11 indices:  0 records:    1 events:        20 fr: 12.9%

                       

                      recno:     12 id:      12 indices:  1 records:    7 events:       996 fr: 85.8%

                       

                      This will at least give you an idea which points are the "busiest" in terms of archiving data. In the snippet above, it looks like the point with pointid = 4 has the most archived events.

                        • Re: PI Point Statistics

                          Hello Mike,

                           

                          I concur with David. All you need to identify what tags require being reviewed is pidiag.exe and a recent not registered archive. I suggest using the 2 most recent files from PI Backup - the primary and the previous primary archive. This way you don't need to unregister. 

                           

                          It's possible to dump the Archive Check statistics into text files:

                           

                          pidiag -archk <path to archive file> > <output file> 

                           

                          With PI Archives we differ between Primary, Overflow and Index records. Index records are pointer records that are used to indicate what Overflow record(s) will contain the queried data. The archive check statistics indicate the amount of Index records per tag:

                           

                          recno:      4 id:       4 indices:  8 records: 1005 events:    493550 fr: 99.4%

                           

                          Because of performance reasons we recommend not having more than 2 Index records per tag. I personally would not consider 8 Index records an indication for a serious issue but at least an indicator that Exception and Compression setting of the tag with pointid=4 should be reviewed.

                           

                          ---------------------
                          edit: added a missing "not"