There is ample documentation in the public domain on selecting and sizing UPS (Uninterruptible Power Supply) capacity for the load it will carry; this article assumes you have already selected and sized yours appropriately, and focuses on a couple of considerations that many small and medium-sized businesses overlook to their peril.
Hard UPS Lessons from the Field
A Clean Install with Essential Loads Only Please… Your Critical Infrastructure should be physically and electronically secured, and installed in a clean and organized fashion. Don’t use power bars with rocker switches off your UPS tangled in cords under a desk feeding critical loads. If your business must have your UPS and server sitting next to an office desk unsecured, ensure the occupant of the office knows not to plug anything else into it, and cannot easily turn the UPS off by mistake. More than once we have seen customers plug in space heaters, laser printers, hot plates, and other heavy electrical loads into their UPS systems, with predictable results.
All windows-based servers need to be shutdown properly, every time.
This is especially true on servers making extensive use of delayed-write disk caching. Your UPS needs to communicate its’ status with your server, and your server needs to be configured to shutdown when running on battery, before the battery is too depleted to allow this to happen. Most new UPS systems will come with a USB connection for a server or workstation to facilitate this; higher-end models will have an Ethernet interface with more advanced management options (such as web and/or SNMP interfaces, and possibly vendor-specific software to monitor and react to UPS Status.) Ensure you get and configure the capabilities you need, and your server power management settings are appropriate, including:
-Shut down the server before the batteries are depleted, rather than Hibernate or Sleep (neither of which are appropriate for any file server).
-Configure power settings in BIOS and Windows to ensure the server does not start up automatically after a power loss; rather, it must remain shut down until someone manually starts it back up, after ensuring the power is stable (as much as possible). This includes ensuring “Wake Timers” such as scheduled tasks, or “Wake On Lan” (WOL) are not permitted to trigger a startup.
The first few minutes to half-hour or so after restoration of electrical service due to a large-scale power outage is precisely when your unprotected equipment is most at-risk, and the quality of that power is likely to be at its’ worst. Often we see the AC supply come on and off repeatedly in this period, with voltage spikes and sags, while the utility is working to re-stabilize the system. If your operations require your servers to be started again as soon as possible, then leverage software monitoring and/or scripting to send administrator notifications to appropriate personnel, for any UPS status changes, and ensure they respond to the site to do so. Have documented response procedures for this scenario, which include contacting your electrical utility provider to ensure system stability prior to powering-up equipment again.
UPS batteries degrade over time
So if the unit has been used for years without battery replacement, don’t rely too much on the estimated run-time shown on the UPS, unless your UPS is a high-end model that features automatic battery condition monitoring and periodic testing (for SMB terms, we will assume here that it isn’t).
Setting your server to shutdown when it hits 10% battery capacity may work great when the unit is new, but 7 years later, you may watch the UPS early during an outage only to find it saying it has 30 minutes of runtime one minute, and then abruptly crashes without any warning a moment later. At minimum, as your unit ages, be sure to set your automatic shutdown to occur with an increasingly-conservative amount of “capacity” left in reserve. Addition of applications and configuration changes can increase the time required for a server to shutdown; if your UPS is small to start with, or the battery is too depleted before the shutdown is initiated, the server may not have enough time to complete the shutdown. Don’t consider your UPS a “Run-to-Failure” item.
Be sure to do your homework, or thoroughly consult a trusted service provider, when it comes to protecting your Critical Infrastructure. As one part of your Business Continuity/Disaster Recovery Planning processes, your UPS investment decisions should be made with full consideration of not only the initial capital cost, but also the probability of power quality or outage issues in your area, and the potential costs of inadequate protection to your business when these occur. If important considerations are overlooked, your file server(s) can literally go up in smoke - along with the years of critical data and services your business relies on – within minutes of that one, snow-laden tree falling on the wrong power line.