Historical data retrieval

Client applications can request that data be retrieved from the historical data server in several ways, including:

• Read raw samples – This provides access to the individual history samples stored in the historical data sets. If data compression is enabled, the number of samples stored per unit of time could vary a great deal. The number of raw samples available online for a history tag that has a frequently changing value and with little or no compression could be quite large.
• Read processed data – This tells the server to return information for a history tag, derived by processing the raw history samples; it usually involves a substantial reduction in the volume of data returned, but the data is still useful for a specific purpose: drawing charts, performing analysis, looking for missing data, etc. Processing involves:
  1. Subdividing the time span bounded by the start and end time into one or more, evenly sized subintervals; for example, 24-hour subintervals for the last month.
  2. Computing one or more aggregate functions for each subinterval.

In either type of request, a time span, ultimately resolved as a start time and an end time, is specified in the request to establish the time boundaries for the request. Time stamps that exactly match the start time are considered within the time span. Time stamps that exactly match the end time are not within the time span.

Start and end times from a client application will often not exactly hit the time stamps for raw samples stored in the historical data server, so the following options exist for specifying how the server should handle the boundary sample situations:

• No boundary sample – specifies that no samples should be returned/used outside of the start and end times. The first sample to be used would be the first raw sample stored after the start time, and the last one used would be the last raw sample stored before the end time
• Outside of span – specifies that the time span should be increased to include the first raw sample beyond the specified start and/or end time
• Interpolate – specifies that a sample should be derived for the start and/or end time by interpolation, as appropriate for the data type, from the raw samples that are immediately before and after the boundary time specified.

A complicating factor for historical data retrieval is that there may be discontinuities (holes) in the historical data record. Reading raw samples should return hole start samples that indicate the starting time and the reason for the hole; for example, the parameter cannot be read, the historian shut down, the tag was not configured to collect at that time, etc. Reading processed data should return sufficient information to let the client know if the requested aggregate was adversely affected by any holes in the historical data record within each subinterval.

Another kind of discontinuity in the historical data record involves requesting historical data for time spans up to and including the present. It is usually convenient for clients if the historical data server would, in this case, return a raw history sample representing what the historical data server currently considers now. It is expected that a client repeating the same request for a now sample some time later would see different, more current, results.