Storing and Accessing “Future Data” in the PI System

Version 1

    This document is an excerpt from the Visualizing PI System Data Workbook v2017


    What is Future Data? (Note: PI Coresight is now renamed as "PI Vision"):


    Many businesses rely on the use of forecast data to predict resource requirements or maintenance activities, find differences between predicted and actual production yields, and so on.

    Prior to version 2015, Data Archive only supported data in real time, not the data from forecast or predictions with a timestamp beyond the current time (i.e. the “Future Data”). With Data Archive 2015, however, this type of time-series data is differentiated from future data giving users the capability of storing and accessing future data. For the two types of data combined, Data Archive 2015 allows storage and retrieval of data with time stamps within the range of January, 1970 through January, 2038.


    How is future data managed differently from historical data?

    Data Archive differentiates future data from traditional real-time data by the newly available PI Point attribute of “future”. This attribute is enabled for the future data PI Point. The future attribute cannot be modified after the PI Point has been created. Therefore, existing historical PI Points cannot be converted to future PI Points.


    To store future data, Data Archive uses separate archives called “future archives” that are created automatically. This is in contrast with the traditional archives used to store time-series data referred to as “historical archives”. Future archives have pre-determined time ranges and are created only when data is received. Every future archive has an initial size of 1 MB, grows dynamically, and has a time range always bound to one calendar month. For example, if a new PI value comes in on December 7th at 09:00 AM and an archive file does not already exist for the month of December, Data Archive creates one automatically. Historical and future archives can be managed independently based on specific needs for data retention, availability, performance, and reliability.


    Choosing between historical or future PI Points is a key decision that depends on whether the data that must be stored is real-time data, that is, from sensors collecting continuous measurements, or data that may not be close to the current time or may be frequently revised (for example, forecasts or predictions). Such critical distinction in stored values is unlikely to change in the life of a PI Point.


    Note: Any historical, non-future, PI Points will reject any data with time stamps that are greater than 10 minutes beyond current time.


    Can future data be accessed by PI tools?


    Typically, future data is generated over a specific time range, for example, a day or week ahead of the current time, and is periodically refreshed when a new set of predictions becomes available. The data forecasts stored in PI can be compared against actual measurements either visually (for example, using graphical PI trends) or analytically (for example, using PI DataLink spreadsheets). Preserving the history of your forecasts may also be useful for model optimizations and “what-if” analyses.