Over the past year, I've seen an alarming increase in the number of data center administrators whose facilities have experienced power problems. I'm not talking about component failures, but rather running out of power to handle the load necessary for their IT infrastructure. Some catch the problem before it turns into a disaster, while others experience data center power outages because of overloading.
But power disasters like this don't need to happen. Let's talk about what IT administrators can do to monitor data center power growth and stave off power demand.
Data center power monitoring
Various products on the market allow administrators to monitor the load of each circuit in their data center. This capability is built into modern power distribution units (PDUs), but surprisingly, many are not monitored. And if you are an unlucky administrator running PDUs that are 30-plus years old and lack monitoring capabilities, you can still determine power consumption in your data center.
This brings me back to my first data center job at a financial firm. In 1979, I landed a summer job at a data center just after I finished my freshman year of college. I installed and maintained the data communications equipment. I also monitored the data center power draw with a Fluke clamp-on ammeter. I would go to each distribution panel, make readings of each circuit and write them down. Twice a month, I typed up a report for the data center master electrician, and the data center used this information to forecast power demand growth. Back then, the expensive components were the uninterruptible power supplies (UPSs). The data center manager needed to minimize use of UPS capacity, so we were careful about what went onto UPS systems as opposed to the non-UPS, generator-backed equipment.
If you have old equipment that you cannot justify upgrading to modern PDUs with integrated power load monitoring, a clamp-on ammeter is an alternative to track your data center's power consumption trends. All data centers should have at least two of these tools. Regular power draw readings should be taken to calculate power demand trends. You should also make it a habit to check the power draw on a circuit before plugging in a replacement or new piece of equipment. This ensures the circuit doesn't accidentally overload.
Remember that the circuit rating of the supply to a PDU is often less than the sum of all the branch circuits going out of it. This means that the current supply to the PDU should also be monitored regularly.
As one of my co-workers says, "You don't know what you don't measure." You must have a process in place to monitor power consumption trends of your data center. Without those metrics, you are driving blind and heading for a data center power disaster.
When power demands exceed supply
So you monitor your data center power demands regularly, have forecasted the power requirement and have taken this information to executives to justify a new data center or major retrofit of an existing facility. But the economic downturn put the skids on those large capital expense plans. What do you do? First, the economy has likely stalled your IT equipment expansion plans as well. If your data center power demands have stopped growing, breathe a sigh of relief, but don't rest on your laurels. The economy will pick up again, and you will eventually need to respond to these demands. Plan for the future now by ensuring you are protected from a data center power disaster. Here are some relatively easy steps to prevent take preventative measures:
- Adopt virtualization. The choices are rapidly expanding and the cost is decreasing. In the commodity server platform market, solutions from Microsoft and Citrix are giving VMware competition. For the small and medium-sized business market, Virtual Iron Software offers a compelling virtual infrastructure management product at an affordable price. IT organizations that have aggressively adopted virtualization have staved off data center expansion projects for an additional four years -- and that was during rapid growth times.
- Consider colocation facilities to handle data center growth. I talked to a large utility supplier that had forecasted when it would run out of power. It had already adopted virtualization and worked out the time required to build another data center so it wouldn't face a power disaster. But the company's executives denied funding for a new data center -- something the IT staff hadn't considered. I recommended that the company handle its expansion by outsourcing to colocation facilities. The colocation trend continues to grow, with new colocation data center construction being near the top of all data center construction projects.
- Decommission older servers. I talked to a large manufacturer that was nearing its data center capacity, but it had acquired another facility into which it planned to expand. The company began its inventory of equipment only to discover that about 25% of the equipment had work orders to be decommissioned. Work orders to decommission IT equipment were the lowest priority within its IT operations and some had sat unfulfilled for nearly 18 months. Server decommissioning is an easy way to recoup space and power -- and extra equipment for lower-tier workloads.
The best way to avoid power disasters is to put in place metrics to monitor the power growth trends at all points in the power distribution path. Plan early to implement interventions to reduce data center power growth, such as virtualization and decommissioning of older equipment, and consider colocation facilities for when expansion becomes necessary.
ABOUT THE AUTHOR: Richard Jones is the vice president and service director of the Data Center Strategies practice at Midvale, Utah-based Burton Group. He can be reached at [email protected].
What did you think of this feature? Write to SearchDataCenter.com's Matt Stansberry about your data center concerns at [email protected].