Trend Sampling Mode

The Trend Sampling mode maximizes performance when retrieving data specifically for plotting.

The Trend Sampling mode identifies significant points and returns them to the caller. These will be raw samples. Significant points are established by finding the raw minimum and raw maximum values within each interval. Note that this is not the same as finding the change in slope direction of a line, as archive compression does.

The Trend Sampling mode approximates a high resolution trend with only as much detail as could be drawn on the page. For example, say you are about to draw a trend on the page and you know that the area with the trend graph is only 100 pixels wide. You could not possibly represent any more than 100 points in those 100 pixels. By using the Trend sampling type, you can ensure you retrieve adjacent highs and lows to draw a visually accurate trend with only 100 points, regardless of whether the time period was one year or one hour.

Since the Trend Sampling mode does not need to acquire all data between the specified start and end times, it is a very efficient method of data retrieval, especially for large data sets. Depending on the requested start and end times (and the amount of data stored for that interval), it could be as much as 100 times faster than other methods.

When displaying data for reports or examination, this principle can be applied to other sampling types too. It is highly inefficient to trend data at a higher resolution than can be drawn on page or printed on hard copy. This is useful even for calculation modes like "Average value". It is not suitable for Interpolated mode, since this results in a loss of detail that ihTrend sampling attempts to recapture.

The ihTrend sampling type returns adjacent highs and lows within each interval. If you ask for 100 samples, you will effectively receive 50 high values and 50 low values over 100 intervals. The retrieval process works as follows:

1. Divide the query duration into even-length intervals, like other sampling modes.
2. Determine the raw minimum and raw maximum for each interval. If there is only one point, then that is both the minimum and maximum.
3. Since we want to return 2 samples per interval (a minimum and a maximum), we need twice as many intervals. Divide each interval in half. For example, a one hour interval of 01:00:00 to 02:00:00 becomes 2 intervals (01:00:00 to 01:30:00) and (01:30:00 to 02:00:00) .
4. Put the minimum in one half-interval and the maximum in the other. If minimum comes before maximum, put the minimum in the first half-interval and the maximum in second half-interval, and vice versa.

When doing filtered data queries, your maximum returned intervals must pass the throttle, even if only a few intervals actually match the filtered criteria.

Timestamp

There is no difference between full-interval timestamps and half-interval timestamps. Both are valid and all interval timestamps are in ascending order.

The Trend Sampling mode will always have an even number of samples, rounded up when necessary.

For example, if you request num samples = 7 or num samples = 8, you will get 8 samples.

If you request results by interval instead of number of samples, you will get back twice the number of results you expect.

For example, a 5-minute interval for a 40-minute duration is normally 40 / 5 = 8 samples. But with trend sampling, you get 16 evenly-spaced intervals.

Value

The raw minimum or raw maximum of the full interval. There is no indication as to which one you are getting.

Data Quality

Trend sampling uses the same logic as interpolated sampling to determine the percent good quality.

select timestamp, value, quality from ihrawdata where samplingmode=trend and timestamp >= 
'29-Mar-2002 13:50' and timestamp <= '29-Mar-2002 14:30' and tagname  = tag1 and numberofsamples = 8

select timestamp, value, quality from ihrawdata where samplingmode=lab and timestamp >= 
 '29-Mar-2002 13:50' and timestamp <= '29-Mar-2002 14:30' and tagname  = tag1 and numberofsamples = 8

[Data] 
Tagname,TimeStamp,Value 
Dfloattag5,01/05/03 8:00,95.00
Dfloattag5,01/05/03 15:00,88.00
Dfloattag5,01/05/03 16:00,80.00
Dfloattag5,01/06/03 7:00,11.00
Dfloattag5,01/06/03 10:00,13.00
Dfloattag5,01/06/03 13:00,93.00
Dfloattag5,01/07/03 8:00,99.0
Dfloattag5,01/07/03 11:00,86.0
Dfloattag5,01/07/03 12:00,16.0
Dfloattag5,01/08/03 8:00,0.00
Dfloattag5,01/08/03 12:00,99.00
Dfloattag5,01/08/03 14:00,100.00

Select timestamp,tagname,value Quality from ihrawdata where tagname =dfloattag5
And samplingmode= trend and intervalmilliseconds =24h
And timestamp> ‘1/02/2003 07:00:00’ and timestamp<= ‘01/10/2003 12:00:00’

Dfloattag5,01/06/03 13:00,93.00

Dfloattag5,01/06/03 10:00,13.00