Servers have long had certain types of environmental monitoring built into them. Administrators can configure management software to detect when the fan dies or the temperature soars and issue alerts accordingly. These days, however, that isn’t enough. Why wait until a server starts melting before discovering a climate control issue?
With environmental sensors priced lower than the cost of a single downtime incident, such devices are a worthwhile investment for the data center looking to keep the disaster recovery plan in the drawer.
The bottom line is you have to monitor the entire server room.
Environmental sensors are not important at all until suddenly they become very important.
|
And that’s where environmental sensors come in. With features like TCP/IP-based remote monitoring, abnormal status notification via e-mail messaging or SMS, and SNMP support, they provide an additional means of safeguarding sensitive resources.
“Environmental sensors are not important at all until suddenly they become very important,” said Jon Collins, principal analyst for Quocirca, a U.K.-based industry analyst firm. “Many IT managers view them as a relatively cheap addition to the server room compared to the cost of downtime.”
Units are mounted on a wall or equipment rack, or placed on a shelf. They can be centrally monitored from a workstation, over the Internet using a Web interface, or wirelessly using a PDA. The devices come with their own specialized monitoring and management software, but since they use SNMP they can also be incorporated into other network management software as another node to monitor. Generally, the base unit of an environmental sensor can be attached to additional sensors as required.
Making Sense of Environmental Sensors
Sensors should monitor for the following:
Temperature: It is not good enough simply to nail a thermostat to the wall. Since the temperature can vary drastically around different pieces of equipment, consider placing separate temperature probes on individual racks or critical devices. That way, problems with a broken fan or air conditioning failure will show up quickly. For best results, stick with microprocessor-based temperature sensors. A CAT 5 cable, for example, ensures accuracy does not diminish due to cabling factors. Avoid sensors that demand calibration to maintain the correctness of readings.
Humidity: If humidity is too high, it can lead to corrosion; if it’s too low, you have static electricity problems.
Airflow: While the building engineering team may be alerted when the air conditioner trips a breaker, IT managers want to be sure the air is actually flowing at a high enough velocity down through the racks. Airflow hovering around the ceiling grille isn’t good enough. The air must get down to the floor-level servers in the path of the air stream where the status and amount of the flowing air can be properly monitored.
Water: Sensors should be capable of detecting the presence of water so remedial action can be taken before it shorts out equipment. Water detectors should be microprocessor based, capable of detecting distilled water, and encased in epoxy so the device can function while submerged.
Voltage: Voltage sensors detect the presence or the absence of line voltage. They can prove useful in identifying the frequency of brown outs for measuring uninterruptable power supplies and service provider performance. The best voltage sensors come with a measurement range between 50v to 250v.
Dry Contact: Dry contact sensors detect broken surfaces, such as when someone opens a door or breaks a window. As soon as someone opens the door to the data center, a trap or page is transmitted informing the admin that someone has opened the door to a secure area.
Sensors should have, at a minimum, the following functionality:
Notification: The IS organization should be able to set temperature thresholds for alerts. For example, if the air conditioner breaks, the temperature sensor should immediately tell the probe to page an administrator. Probes should be able to send out alarms via e-mail, SMS, or SNMP trap.
Interfacing: Environmental probes should be able to interface with the major SNMP-based network management systems, such as HP OpenView, IBM Tivoli, and CA Unicenter.
Camera: Cameras add an extra layer of security, as they enable staff members to see who is in the server room or remote server closets. They also provide the IS organization with a record of who was in the various rooms and what they did there. Cameras are also helpful for remote offices without a local IT rep. When troublshooting issues, for example, someone can walk around the server room with a camera while on the phone with the main data center.
Today, some probes come with cameras that have adjustable mounted lenses with widescreen pan and scan features. Be aware, though, that cameras can add significantly to the price tag. We recommend adding one only if you are fairly certain you willuse the features.