Communication services · 24 June 2016
Polling Going Deeper
The very basic polling service described in the first part has a limited throughput, so it is not well suited for real applications (except for very small ones) in its original form. Could it be enhanced? Of course.
Divide and conquer
Even a big system could be divided in small parts. Every part should have own master (router / gateway / main controller). This makes a system more complex, expensive, harder to maintain and more error-prone. Still, quite often building management systems are designed with one gateway for every 5-10 field devices like energy meters.
Speed it up
Increasing the baud rate is the first thing that comes to mind. Most serial devices go higher than 9.6 kbps and support 19.2 kbps, 38.4 kbps or even higher speeds. 19.2 kbps is the Modbus default, Sauter novaNet uses 19.2 kbps, BACnet MS/TP – up to 76.8 kbps, LonWorks FT-10 – 78.1 kbps.
But it doesn’t come free – when the speed is increased, device CPU should be more powerful, cable maximum distance shorter, noise sensitivity higher, requirements for cable quality stricter and termination resistors become necessary. Which leads to extra costs in materials and commissioning labor for troubleshooting. That’s why many engineers and manufacturers prefer to keep the baud rate low: Grundfos GENIbus and Satchwell Micronet NCP operate at 9.6 kbps.
Get what you need
This is an obvious idea: read only the information which is needed at this moment. As graphical user interface displays only a limited number of points at once, only those should be polled. There are a few things, though:
- Quite often point values are used not just for displaying, but also for alarms or trendlogs – then point values are needed all the time
- Software design is much simpler if all points are polled continuously, therefore manufacturers often choose simple route. Especially in case of protocol gateways
Tridium Niagara deals with this aspect very nicely. A point is read from a device only when it is subscribed. And it happens when either:
- point value is visible on user screen
- point has an extension – alarm or history
- point is linked to some logical component
- point is leased (temporarily subscribed) by a program
This way only necessary points get polled and the bandwidth is used efficiently. All Niagara drivers inherit this functionality automatically, which is very useful.
But then again, Niagara is (probably) the most sophisticated software on the BMS market and many other engineers have to rely on gateways and less advanced software, which often lack similar functionality.
What other options are available to increase the point update rate? Point grouping.