It can happen easily: Someone opens the back door to haul freight into a data center during a thunderstorm, and the warm air hits a chilled row of racks and storage gear. Moisture condenses and systems errors ensue.
Environmental monitoring systems for data centers and server rooms can help IT professionals mitigate downtime. Available from makers such as APC, Liebert and Raritan, the software sends alerts when thresholds are exceeded for criteria such as temperature, humidity and open rack doors.
These systems can’t be effective, though, unless someone receives the alarm and acts on it. Thresholds need to be set accurately and tested. And sensors can go bad, become disconnected or surreptitiously be cut. For instance, one data center had a chain of sensors fail simply because a primary door that was frequently opened and closed wore through the connecting cable. So it makes sense to test sensors too.
What follows are some tips for best deploying environmental monitoring solutions in data centers, server rooms and wiring closets.
Manufacturers of hardware chilling components and racks usually supply monitoring apps. Some of these rely on connectivity among sensors, while others can ride on Ethernet cabling that’s part of the management network. Choose an app that meshes well with your sensors and other management platforms.
The debate about using Ethernet infrastructure rather than bus-like telecom cabling for alarm data transport has raged for years — one argument is that if Ethernet goes down, an out-of-network connection scheme will still work. Others maintain that what kills one, kills the other.
Route the sensor circuits in the manner recommended by the manufacturer plus all of the Layer 1–3 circuitry to where the management app resides. Consider using Ethernet virtual LANs to aggregate management traffic away from production traffic with switch port control.
Regardless of which alarm data transport method is used, the alarms must still be communicated to a working console, which in turn must get the attention of someone. A monitoring app worth its salt must be able to send and receive email and texts, and trigger IPMI, SNMP or other correctly configured-and-tested console alerts to get the attention of someone who can deal with the alarm in an appropriate way. The best console is worthless if it can’t gain someone’s attention to resolve alarms.
Some monitoring apps can write to either Microsoft event or syslog files, which goes a long way toward log management of events. Both formats have a hierarchy of urgency as to the type of message. Did you care that Rack 49 in Room 105 was opened for three hours, or do you care more about the subfloor water detector underneath Rack 49? This is why message urgency makes a huge difference in outcomes.
Qualifying an app for your circumstances means making sure that it has the ability to talk to your existing management apps or provides the protocol access that delivers the goods. Vendor-specific apps can be enormously flexible, but they can’t live in a bubble. When all else is said and done, relieve late-night boredom by holding test drills to ensure that the alarms can still trigger a notification that the barn door is open.