As I was sitting at my desk deciding on a topic for this month’s blog, I began pondering what cool exciting new network gadget I could review or elaborate on. Writing about new technologies and discussing their impact upon our organizations and lives usually allots for topical discussion. As I was brain storming, a friend of mine called me and began reiterating about the stressful drama he was forced to endure the previous day at his job where he manages a data center for a school system. The area had been ravaged by violent storms the night before, resulting in numerous power interruptions throughout the day. These power fluctuations ended up wreaking havoc on his datacenter. The casualties the next morning were not pretty. Three power supplies were deemed inoperable, one of them evoking a burning smell throughout the datacenter and a power strip literally burned out. On top of that, all four clustered Hyper-V hosts eventually lost power, resulting in the temporary outage of over thirty virtual servers. The primary SAN lost power that resulted in a cache corruption which would require hours of troubleshooting. His disturbing tale was a complete spectacle of excitement, too much excitement, which is why I decided to write about a very lackluster topic, Uninterruptable Power Supplies. It is my intention to bore you with this humdrum subject matter so that you can avoid the agitation that is inevitable if one chooses to avoid this product category that few give proper credence to until it is too late.
If you think that power disruptions to your datacenter are a low risk, think again. In a recent study of 584 data center operations personnel, 85% of respondents reported having experienced a loss of primary utility power in the past 24 months, and 91% of those said they’d had unplanned outages. In fact, respondents on average reported two complete data center shutdowns during the preceding two years, with an average duration of 91 minutes. Couple that with the fact that only 38% of respondents said they have ample resources to bring the data center back up in case of an unplanned outage.
Downtime is expensive. The national retailer, Sears, reported that a five-hour failure of four USP devices at the start of 2013 cost the company 1.58 million dollars in profit. Problems continued well after the event and the company was forced to run generators for eight days to keep the server farm running which cost an additional $189,000.
Boring sounds pretty good about now doesn’t it?
A UPS device should be something that you put some time and effort into the purchase decision. The largest manufacturer of UPS devices is APC which offers a plethora of products offerings covering every need and price point. When deciding upon a UPS, consider these important tips:
- A Warrantee – A UPS promises to protect equipment worth thousands, perhaps tens of thousands of dollars. In a sense, you are purchasing “peace of mind” with a UPS and legitimate manufacturers reinforce that through a connected equipment protection warranty. Terms of the warranties can vary, but most will offer both a product warranty (which covers defects in the UPS itself) and a “connected equipment” warranty, which promises to reimburse the value of any devices destroyed by failure of the UPS. The value of the connected equipment warranty can vary from the tens of thousands to hundreds of thousands.
- Plan for Enough Outlets – Make sure you know how many outlets you need. UPS devices are commonly associated with servers and the need to keep them powered up long enough to conduct an orderly shutdown. Don’t forget thought about your other datacenter devices such as switches and routers. For instance, you may have purchased a cloud service that provides cloud server redundancy just for situations such as this, but unless your routing and/or switch devices remain online, your off-premise resources may be rendered as useless as your on premise resources. Also, on cheaper UPS models, don’t assume that every outlet is powered by the battery. Read your documentation to make sure. Also, if your datacenter utilizes VoIP for its phone service, you must have these resources protected as well so that your datacenter personnel can remain in communication during a power outage.
- How Long Do I Need to Keep my Equipment Up For – At the minimum, your datacenter personnel need enough time to conduct orderly shutdowns of any servers or storage devices such as a SAN in the event of an outage. You need to realistically decide how much time would be necessary in such an event. For those organizations fortunate enough to have a backup generator that is powered by a gas or fuel line, you may need just enough time for the generators to kick in. If you do have backup generators, remember the principles of redundancy and don’t put all of your eggs in one basket. It has been my experience that backup generators don’t always get triggered during brownouts.
Next month we will cover a few more tips and then look at the various IPS topologies available today and the advantages and disadvantages of them and then explore some of the advanced features that you may not be aware of.
bradinthecloud 
Posted on May 14, 2014
0