We are working on a project to monitor the power in an entire datacenter (around 300 cabinets) and I’m looking into the emonTx for that (installed on the top of the cabinet). I have no experience with devices working at the 433MHz frequency. Will it be a problem to have a very large number of devices using this frequency at the same location?
The datacenter has multiple rooms and we’d deploy different wireless bases for each one, but even so the number os devices connected to each base would be around 50 or so.
You are going to struggle with having that many RF nodes. For starters there are only 30 unique node id’s available on each network group, and even if you have multiple network groups, they will still cause congestion as each packet needs to be received before it can be processed or discarded as it belongs to a different group.
If you could use 2 different frequencies in the same room, eg 433MHz and 868MHz (said without knowledge of whats allowed in your locality) you might be able to get all ~50 recognized in theory, but unless you synchronize the transmissions you will still get clashes that result in data lost unless you slow the transmissions right down and keep them short.
And, the other rooms (and other electronics) may cause further interference.
In reality the RFM is best for a small number of nodes, perhaps a maximum of 10, I’ve noticed data loss with just 3 emonTx’s when @10s updates and that’s without any other nearby groups to cause further issues.
I would recommend using some FDTI adapters and connect them by usb (in batches using hubs) to a emonBase Pi per room, additional expense I know, but money well spent considering you will get greater granularity updates, no data lost and no time debugging and messing with RF issues.
So maybe I’ll just stick with a the regular shield stacked with an Ethernet one and do it all over IP and SNMP. This is even easier because we can also provide DC to power each probe over the existing Cat6 cables.
Yes the commercial solutions are definitely way more expensive, I have no idea of the going rate for an Ethernet shield as I have never used one, I might take a look into that option myself one day so please keep us posted on your progress.
If you are running a commercial data centre, I’d avoid the Arduino ethernet shield - they will be wiped out as soon as the network has some sort of security scan/pen test done against them. They are not robust to cope with high volume traffic and/or hacking attacks.
Cabled serial data is probably the best solution as already mentioned - perhaps using a shared serial protocol instead like RS485 ?
If done right, I don’t see a reason why this wouldn’t work. We do commercial power rack monitoring for data centers, UPS, generators, etc. as well as HVAC control of pressure, temps, humidity etc. In my experience we always metered at the sub panel, the circuit(s) that the rack is getting power from something like a PMM (Power Management Module) and installed this
I have a set like this installed at my house and am using emoncms to collect/analyze the data.
Instead of getting a bunch of emonTX’s to monitor, installing a set of Veris H704 Ct’s at each of your electrical panels is the way to go.
These current sensors are sensing in mA and can be converted to Amps, kW, kWh, etc inside emoncms.
Emoncms platform is far superior than that of a commercial grade software that is available, emoncms takes the cake in terms of broad array of processors, trends, graphs, reports. etc.
How is your one line electric diagram power connected? can you sketch a rough drawing of how the racks are powered?. I’m assuming they go to a distribution panel somewhere.
I know the $ is a bit much than that of a emonTX, but with the Veris hardware you are monitoring several racks. It all depends on how the racks are powered, and if one panel is supplying several racks with power. A sketch of how the data center power is connected would help alot.
You can only get apparent power by multiplying voltage and current in emoncms, that is only close to accurate when monitoring loads with a near unity power factor. I’m not sure servers are likely to achieve a good enough power factor to get an accurate indication of real power or energy consumption.
With the kind of servers and network equipment we have today the power factor is very constant on Internet datacenters. If we get a customer with something different (a big storage with lots of disks, etc) we do a manual measurement of kWh with a special device and correct the power factor calculation for individual cabinets.