13 Replies Latest reply on May 4, 2010 7:05 AM by MichaelvdV@Atos

    The new PI / PISDK Engineering plans

    MichaelvdV@Atos

      I think a lot of us are waiting in anticipation for PISDK 1.4.0. It was first planned for Q1 2010, then for Q2 2010, and now in the PISDK Release Schedule it's planned for H2 2010 (which could eventually mean it will be released as early as July or as late as December 2010.

       

      PISDK 1.4.0 would mean that we can make real (and easy) use the HA possibilities of a PI server setup. It seems a very essential piece of the puzzle.

       

      I was wondering if anyone could comment on the reason of the delay, and maybe has some more information on the actual release time frame.

       

      Another issue is the PI Enterprise Server Engineering Plan. Not so long ago, the SSB feature was mentioned for a future release. Some time before that, Future Data was mentioned for a future release. The current Engineering Plan does not mention both functionalities.

       

      We (as a company) are in dire need for Future Data support, and are now implementing it using the OLEDB COM Connector. Since this is not a very good long term solution for high demand systems, it would be nice to have an idea when we can expect to upgrade to a PI version that natively supports Future Data. Can anyone comment on this?

        • Re: The new PI / PISDK Engineering plans
          mheere

          HI Michael,

           

          I can't speak to the timelines, but I do have two questions about your scenario:

           

          1.  I see that you participated in this other thread: http://vcampus.osisoft.com/forums/p/148/530.aspx#530 about future data.  Can you clarify for me whether you have only the future timestamps issue, or if you also have the data versioning issue?

           

          2.  Can you elaborate on what functionality you're looking for to be provided by SSB?  In particular, what is it that you need that PI to PI does not provide?

           

          Thanks,

           

          Matt

            • Re: The new PI / PISDK Engineering plans
              MichaelvdV@Atos

              Thanks for responding.

               

              At this point, it is only the future timestamp issue (and all the problems that come with it). Versioning is not an issue (at this moment). Only the most recent forecasts (and in our case 'orders') are stored. The older values are deleted. At this point, we are using the PI OLEDB COM connector to store the data in SQL server using COM connector tags.

               

              The main issues with the PI OLEDB COM connector are:

              • Performance is very weak (compared to native PI tags)
              • Single Threaded
              • No update functionality (cannot update an event)
              • No delete functionality (cannot delete an event)
              • No annotations
              • No Digital States for non-digital state tags

              If anyone has more experience with the COM connector and knows an elegant solution to any of this issues, please let me know!

               

              So, as yo can see, the COM connector tags are not fully 'transparent', in the way that they don't offer the functionality that a native PI tag does. This is a real issue. We are talking about a system with about 10k - 15k Future Data tags. This could become an issue with using AF for frequent calculations.

               

              SSB was mentioned as an example of functionality that was stated in the Engineering Plan (in the past). I don't need it (I think) that hard right now. Future Data support and PISDK 1.4.0 we really (really, no really!) need

               

               

               

               

                • Re: The new PI / PISDK Engineering plans

                  Michael,

                   

                  I asked similar question outside of vCampus before, the answer I got was that when OSI first mentioned SSB it was a high priority item to address.  Then their priorities changed and it is hard to get it back to their high priorities, recently they switched their attention to AF v2.x and PI Server WIS - they addressed more urgent needs.  I got the impression that securing existing products (e.g. the server itselfs) and delivery the next generation items have higher priority right now...we really need the PMs spin on this as they can describe it better than I can   I do remember (verbally / email ??) that it was mentioned the engineering plans always included lots of far in to the future enhancements that gave the wrong expectations to customers, so now we see very little in to the future on the plans.

                   

                  Steve gave a data access presentation at the UC using PISDK with buffering

                   

                   

                  • Re: The new PI / PISDK Engineering plans
                    Bryan Owen
                    Oddly enough, PI clients may offer functional parity between AF attributes and tags before future data is mainstream. Since the future data sets are already in RDBMS the AF table or custom data reference approach might be viable.  The latter especially if the use case requires bulk forcast query as opposed to adhoc access.  Based on some prior posts, it seems you might have considered AF DR already? Some of the PI_OLEDBCTR limitations can be stretched when implementing stored procedures but it sounds like this use case has pretty much hit the wall. -bryan
                • Re: The new PI / PISDK Engineering plans
                  matthew.rivett

                  We are also looking forward to support for future data.  We store some forecast information into PI using the RDBMS interface.  To do this week have to subtract 7 days from the forecasted date.  So the data shows up as historical data in PI even though it is actually future data.  It can be confusing and misleading to our users who are not familliar with the tags.

                    • Re: The new PI / PISDK Engineering plans
                      Asle Frantzen

                      At Amitec we've also had our share of head scratching on how to accomplish a few future data tasks/issues, forecasts are one of them - as most customers have some sort of predictions they want to save - in order to trend it against the actual production at a later time.

                       

                      (For the customer who only needed a forecast chart series in addition to the current chart series, it was simply solved using a named range from an Excel sheet, which showed up nicely in the future in our RtTrend)

                    • Re: The new PI / PISDK Engineering plans
                      jlakumb

                      Yes, these are much anticipated enhancements for PI SDK and PI Server.  Let me see if I can shed some light on the subject.

                       

                      First, PI SDK 2010 (v1.4) release was delayed due to a change in scope.  We decided to increase the scope to buffer all writes including update, replace, and remove.  Note, most of the changes to support client-side buffering are actually implemented in PIBufSS, so it will be included in SDK.

                       

                      Second, the next major release *after* PI Server 2010 will include Future Data support.  Note, there may be PI Server 2010 R2 release, but Future Data will be included in the release after that (e.g. PI Server 2011 or 2012).  We will update the Engineering Plan when we have a better idea of release date.

                       

                      Regards, Jay

                       

                      PM, PI Server and Data Access

                        • Re: The new PI / PISDK Engineering plans
                          MichaelvdV@Atos

                          Thanks all for responding!

                           

                          As mentioned before, I really appreciate the effort that OSIsoft puts into creating very stable and (sometimes insanely) scalable products.

                           

                          Some thoughts tough:

                           

                          HA has been around for sometime, but the programming tools (PISDK) for supporting this are lacking the proper support. This seems a bit off. It almost seems like 'PISDK is just not there yet'. Can any of you shed some light on the actual release date / approximation of the release date?

                           

                          Next to that, I think forecasting and data in the future is something we will see more and more often as more companies achieve a a level of 'continuous improvement'. As many of the competing products (won't mention any names here before I will be chased with pitchforks by an angry vCampus mob ) do support future data, it is a bit of a shame (and sometimes a real loss) that PI doesn't support it yet.

                          • Re: The new PI / PISDK Engineering plans

                            Jay Lakumb

                            Second, the next major release *after* PI Server 2010 will include Future Data support.

                             

                            Jay, any clues on this?  Will this be stored in SQL Server given the PI + AF bundling now or some other neat tricks with "future archives" type of implementation?  Whilst I am on topic, what ever happened point partitioning?  Not that I have an overwhleming use case, just remember it being mentioned couple of years ago.

                              • Re: The new PI / PISDK Engineering plans
                                jlakumb

                                Yes, future data will be stored in the PI Server as a "future archive".  It is true that we have spoken about many long-term plans at User Conferences (point partitioning, server-side buffering, etc.).  These features are still in our vision, but we are going to try and do a better job of managing customer expectations by sharing what we are actually working on for the near-term (i.e. what you can actually use today/soon to get more value from PI system).

                                 

                                Michael @ Atos Origin

                                HA has been around for sometime, but the programming tools (PISDK) for supporting this are lacking the proper support. This seems a bit off. It almost seems like 'PISDK is just not there yet'. Can any of you shed some light on the actual release date / approximation of the release date?

                                 

                                First, we put a lot of effort into implementing the HA connection management (e.g. failover) in PI SDK.  Sure, one could argue that writing to HA PI is just as important, but we decided to defer that capability.  We are now returning to fill that need with SDK buffering.  Look for that to release in H2 2010 (see Engineering Plan for updates).

                                 

                                Jay

                                  • Re: The new PI / PISDK Engineering plans

                                    Jay, any idea on how far in to the future PI will store data?  Will it be configurable based on the future archive(s) the admin creates/modifies (in PISMT will we see a seperate section for future archive management) or will it be strictly limited to an OSI determined time span?  So many questions...

                                      • Re: The new PI / PISDK Engineering plans
                                        jlakumb

                                        We are not making any assumptions on how far into the future that users will want to store data.  We have heard everything from 24 hours (e.g. load forecast for T&D customers) to decades (scheduled maintenance data for large assets like transformers).  Honestly we are still in the design phase so it is too early to say what the system management experience for future archives will be.  Hope this helps...