Page 261 -
P. 261

232   P a r t   V :    T h e   G r e e n i n g   P r o c e s s



                      the missing 34 percent doesn’t just disappear. It turns into heat—that same heat you have to
                      pay to cool off.
                         When power supply efficiency and system cooling are taken into account, the actual
                      useful work done by the computer can be less than 50 percent of the total power draw of
                      the system. That means over half of the power consumed is simply a drain on the ROI,
                      an increased operating cost with no benefit to the client.
                         There are various power supply calculators available for different machine types and
                      uses. At Link 11-4, we have provided you with a list of several that you may find useful.

                      Loads
                      In terms of efficiency, server power supplies are the least efficient when they are just idling.
                      That doesn’t mean they are using more power, they’re just using power inefficiently. It’s like
                      an automobile. When you’re parked with the engine running, you’re still consuming gas, but
                      not going anywhere. When you pull out of the driveway, you’re finally going somewhere,
                      but now wasted energy is a smaller percentage than it was when you were just idling. Now,
                      consider a server that is configured minimally, but with a large power supply. The power
                      supply is churning out a lot of unused power.
                         Manufacturers are trying to address this issue. However, as you’ve seen before, there
                      will be more upfront costs, especially in the form of more expensive circuit materials.
                      Redundancy
                      Quite often, datacenters are configured with redundant power supplies. The reason is
                      obvious—if one power supply fails, the other is there to keep the server online. In fact, in
                      large datacenters, the power supplies are on separate power grids, so if the AC power goes
                      out on one grid, the other supplies power and the servers keep working.
                         Redundancy is great for the sake of ensuring uptime, but it’s tough on energy efficiency.
                      For example, if a server needs 200 W to operate and it has a single 800 W power supply,
                      then that server is using just 25 percent of the supply’s total capacity. If a redundant power
                      supply is added, power use is split between the two supplies and the draw on each is only
                      100 W. The efficiency then drops from 83 percent for a 25 percent load to 65 percent for
                      a 12.5 percent load.
                         As with many issues in the datacenter, this comes down to a balancing act and some
                      negotiations. How important is system availability versus reduced energy usage (and cost)?
                         The biggest thing you should take from this chapter is the importance of consolidating
                      and virtualizing your servers. The technology now exists to run multiple servers virtually.
                      In the next chapter we’ll talk about server consolidation and show you how you can do it.
   256   257   258   259   260   261   262   263   264   265   266