In the first
of two articles about power quality, we outlined the basic elements needed to plan and implement a data center power project.
In this article, we will review some of the important factors to consider when reviewing vendor proposals. You should take the following into account:
Sole Source -- How valuable is it to have a single point of contact for the whole solution, including wiring, installation, maintenance and so on? Is it worth the effort and potential finger-pointing to go with separate vendors versus one group to hold accountable?
Power Conditioning -- Line conditioners are designed to smooth out spikes, sags and remove EMI/RFI and harmonics from the line. Some UPSes (uninterruptible power supply) have line conditioning as an integral subsystem, while others require an external system to be effective.
Part of the site survey should include issues detected. Note, there are many AC power issues to consider beyond voltage, EMI/RFI and harmonics. A search on Google can readily turn up many sources of additional information. Also, hardware vendors sometimes can provide insight as to what their systems can tolerate.
Lightning Strike Suppression -- Some line conditioners can handle a lightning strike without any problem, some marginally so and others not at all. Make sure that your power systems are capable of protecting your electronics. (A small footnote: You may have other lines going into your building that can be affected by lightning and may need protection. The author lost a fairly expensive PBX once due to a lightning-related power surge that came through an unprotected analog phone line port.
Uninterruptible Power Supply (UPS) -- There are two main types of UPSes, differentiated by how they handle power prior to a power loss. There are units that are online all the time -- thus, all power supplied to the data center flows through the UPS first. Due to this filtering, these systems provide conditioned power and there is no switching time associated with converting from "street" power to UPS power.
The second type of UPS is often termed a "standby" unit, as it is offline until the power fails and then there is time associated with switching from street to UPS power. This switching time may only take milliseconds, but there are some devices that are so sensitive that they can crash during that time.
All things being equal, an online UPS is recommended for critical data center power. In sizing the UPS, take into account how long you want to keep the systems up plus how long it takes to do an orderly shutdown. Make sure that policies and procedures reflect this timing element.
Electric Generators -- There are many types of electric generators. There are ones fueled by gasoline, natural gas, propane, and diesel. Your organization should select the fuel based on what will be most available during a crisis in their area. For example, inline natural gas sounds appealing, but can the gas company be depended on? The generator should be sized to meet the data center's need and expected growth.
Keep in mind, however, that a generator does not replace a UPS. When power first fails, the data center needs to run off the UPSes while the generator(s) start and stabilize. Power is run from the generator through the UPSes to garner conditioning benefits and to recharge the batteries in the event of a generator failure, refueling, etc.
Monitoring & Alerting -- The software and hardware needed to generate alerts when street power goes down and the UPSes and generators go online are cheap investments. Look for hardware and software combinations that allow for logging of power data and alerting when certain conditions are met. In remote data centers or in shops that aren't 24x7 staffed, the alert may be the first indicator that there is a problem. Also, be sure that the alerts can be sent via the Internet, phone and/or wireless allowing for increasing levels of fault tolerance.
Environmental Issues -- Don't forget the air-conditioning! There are to parts to this. First, don't forget to factor in and supply the power required by the air-conditioning units that cool the data center. Second, don't forget to factor in the heat generated by the UPSes on the air-conditioners. Large UPSes generate significant amounts of heat and if they are located in the data center, then the cooling requirements will change and may overwhelm existing AC units. Another environmental issue to take into consideration relates to water. When the UPSes are placed, look at where water pipes and sprinklers are located and consider what they may do to the unit.
Fault Tolerance -- Having redundancy is like how fast a person wants to be able to go in a car. It's a function of money! If the risks warrant it, groups install power coming from separate grids, multiple site UPSes with multiple generators and multiple fuel supplies. The risks and service levels need to drive the level of investment and redundancy.
Expansion/Growth -- When purchasing systems, establish a timeframe of how long the current purchase must be in place. Next, establish a projected power growth rate and error factor. This will help both you and the vendor in terms of sizing the appropriate system. Some solutions are very expandable and others either aren't or have very limited expansion capabilities.
Scheduled Reviews -- Establish a review cycle when the stakeholders will get together both with and without the selected vendor to see how things are going. Power needs change and some environments are more predictable than others. The intent of the meetings is to ensure that the power system is meeting current expectations, projected needs and to identify any corrective actions that may be needed.
As you can see, there are many factors that need to be considered ranging from the risks to the potential solutions. An organization that carefully understands requirements and factors in risks, benefits and costs associated with quality power will be prepared for the next crisis.