Whether it’s to save money or to reduce their carbon footprint, many companies are looking for ways to run more efficiently. The best place to start, according to a recent Gartner webinar, is the data center. Most data centers consume 100 times more energy than the offices they support. A great deal of energy is required to cool the servers and run the facility, of course. But servers also need power to run the applications. The amount of energy needed varies greatly by the type of server and the number of users in the system.
SAP realized this and developed the SAP Server Power Standard Application Benchmark as way to measure application performance in terms of server power consumption. Performance is defined as SAP Application Performance Standard (SAPS), or more specifically: how many order line items can be fully processed in a given amount of time. Power consumption is measured in watts. The test reveals how many watts a server requires to achieve 1,000 SAPS. The lower the number of watts, the more efficient the server is running.
Fujitu’s server now uses 43% less energy
The idea is that customers can use this figure to easily compare the efficiency of different servers before they make a purchase decision. At the moment, however, Fujitsu is the only SAP partner to have participated in the test. It has now performed the benchmark two times on its PRIMERGY servers, which are designed to meet state-of-the-art energy-saving requirements standards. Results from the latest test show that Fujitsu’s newest server is even more efficient than previous versions.
The most recent test from 2012 shows that Fujitsu’s PRIMERGY RX300 S7 server, based on the Intel® Xeon® processor E5-2660, consumed on average 10.5 watts per 1,000 SAPS. Compared to last year’s test, where the S6 server generation consumed 18.3 watts/kSAPS, their new server demonstrates a 43 percent decrease in power consumption.
Less energy needed for all loads
The SAP benchmark measures server performance across nine different load levels in order to simulate the real-life variation in use that a server deals with over time. The 100 percent load level was defined in advance of the test as 6,150 users. This is the maximum number of users that can be operating on the system while still maintaining a response time of less than one second. The other load levels represent a percentage of the maximum number of users (80%, 65%, 50%, 40%, etc.). In addition, the benchmark also tested power consumption during active idle, which is the time when there is only one user on each instance of the SAP solution.
The average power to performance ratio across all nine load levels was 10.5 watts/kSAPS. In itself, this number demonstrates a high amount of energy efficiency, but it’s also worth it to take a look at the results for each individual load level. The test shows that Fujitsu’s new PRIMERGY RX300 S7 server achieves energy efficiency at a much lower load level compared to the older server. Namely, the S7 server starts achieving optimal energy efficiency at around only 40 percent load, while the S6 server doesn’t start becoming efficient until the load is at 65 percent. This means that the new server is able to run highly efficiently for a broader range of workloads, giving a company more flexibility in its system landscape management.
It is also important to note that compared to last year’s test, the average throughput increased in the 2012 benchmark from 11,810 SAPS to 15,480 SAPS. So the PRIMERGY RX300 S7 server not only consumes less energy at each load level, it actually increases performance, too.
Dynamic IT infrastructures are also necessary
Lastly, Fujitsu points out that energy-efficient servers alone cannot guarantee a highly efficient operation of the datacenter. Dynamic IT infrastructures are needed to distribute the load across the IT resources in a smart, efficient way. Fujitsu’s FlexFrame for SAP, for example, is one such solution that flexibly allocates physical and virtual processing power to applications for improved efficiency in the entire datacenter.