KbA0001: Optimizing Your Site
Description
Optimizing your BACnet trend log configurations is important to obtaining the data collection best performance from your CopperCube, and for providing the highest quality data for the analytics performed by Kaizen.
Guidelines for Trend Logs
We offer the following recommendations to help you get the most out of your trend logs. They are based on our many years of experience with BACnet, Delta Controls controllers, and data analytics.
Trend Log Size
Trend Log objects should be sized to accommodate at least 72 hours of data. TL buffer size is a trade-off between controller memory and the Trend Log sample rate. Keeping at least 72 hours in the trend logs balances:
- your need to survive a long weekend of Internet outages without data loss [larger TLs are more tolerant of external outages].
- the controller’s memory limits [larger TL buffers reduce the memory available for other controller functions].
- the collection frequency of the CopperCube is needed to collect the data before the TL ‘ rolls over’ [small TLs need high collection frequencies, which generates a lot of traffic].
- the BACnet network bandwidth was consumed while collecting. [larger TLs reduce traffic through the use of fewer larger packets].
Polling Interval
Controllers can poll data as fast as once every second. This is useful for tracking fast-changing input data – allowing the controller to make quicker response adjustments. However such fine-grained trend log samples quickly overflow TL buffers so very large TL buffers are needed to record any length of time. In turn, then more resources are consumed to collect and process the data. And (perhaps counter-intuitively) such fine-grained samples, make very poor candidates for the type of data analytics that Kaizen performs. Very fast TLs are good for local troubleshooting but very poor for revealing large time-scale patterns; and for comparing with other trend logs.
Therefore we recommend using only the following [simple multiples] TL Polling Intervals of 1min, 5min, 15min, 30min, hourly, and daily. A 15-minute interval is a sufficient resolution for most Energy and HVAC applications.
Note: the CopperCube will refuse to collect TLs using Polling Intervals faster than 1 minute.
Change Of Value (COV)
Change of Value (COV) is a useful trend log option for logging ‘event triggers’ – i.e. data points that seldom change but the time at which they change is important. Many people mistakenly believe that because COV trends are sparsely populated (and thus can span long time periods) they are more efficient than using a polled TL. Quite the opposite is true. COVs place a very heavy processing load upon the controller. Even a moderate number of COVs will adversely affect a controller’s performance. Secondly, COV data is much harder to use in analytics and data comparisons. The storage space gains are very quickly lost in the increased processing penalties.
The worst-case is using COV to log a frequently changing data point. Once a data point is changing more than 300x a day (~i.e. every few minutes on average) then a 1-minute polled TL will provide very similar [timestamp] responses (i.e. 1440 samples per day) at a fraction of the controller and analytics processing load. Disk storage is very inexpensive, so sacrificing storage for improved performance is very worthwhile.
Polled trend logs should be used in most situations. COV should be reserved for exceptional circumstances.
Device Clocks
All clocks on all Devices should be synchronized to a recognized time source.
Clock Synchronized Polling
All Polled TLs must be synchronized to the clock – meaning samples are taken on the hour and every interval thereafter; not at some random minutes+seconds past the hour. For example: a 15min Polled TL shall take samples at 0:00, 0:15, 0:30, and 0:45. An hourly TL shall take its samples on the hour. A 1 minute TL shall take its samples on the minute, every minute. Failure to take samples at the correct times results in data that is very difficult to use, and analytics that can’t deliver their full potential.
Delta controllers have long had this ability built-in. Ensure the TL is configured to use it!
Standard Trend Log Sizes
Adequate Trend Log Size = (72 hrs * 60 minutes / hour) / (Log interval in minutes).
| Interval (minutes) |
Poor [24 hrs] (samples) |
Minimum [72 hrs] (samples) |
Better [1 wk] (samples) |
|---|---|---|---|
| 1 | 1440 | 4320 | —* |
| 5 | 288 | 864 | 2016 |
| 10 | 144 | 432 | 1008 |
| 15 | 96 | 288 | 672 |
| 20 | 72 | 216 | 504 |
| 30 | 48 | 144 | 336 |
| 60 | 24 | 72 | 168 |
*Not possible: Delta controller maximum TL size is 6000 samples.