SPARC T3 Servers Offer an Attractive Platform for Virtualization and Consolidation

(As of Monday, 20 September 2010)

Oracle's SPARC T3 systems enable users to migrate database workloads off older x86 servers to a new SPARC T-Series server. As this benchmark demonstrates, up to 30 x86 servers running Oracle Database 11g can be consolidated to a single SPARC T3-2 server running Oracle Solaris 10 with Oracle Solaris Containers in a secure, virtualized environment.

This consolidation scenario demonstrates a 10x physical footprint reduction, 6.7x better power consumption, and reduced management costs, while providing customers with 20 percent more room for additional growth on the SPARC T3-2 server.

This benchmark was created to the capture benefits of no-cost system consolidation and virtualization opportunities available with SPARC T-Series servers running Oracle Solaris and Oracle Solaris Containers. Older x86 systems typically run legacy applications that are not in high demand In fact, the average CPU utilization of these servers is seldom more than 10 percent. The benchmark uses Oracle's in-house OLTP workload to generate the appropriate load.

In the first part of the test, the x86 servers are tested individually, and data such as CPU utilization, transactions/sec, and power consumption is collected. Then the SPARC T-Series server running Oracle Solaris is configured with the number of Oracle Solaris Containers that correspond to the number of x86 servers. Each container is loaded with the same parameters and workload levels that were recorded on each individual x86 server, thus assuring that demand on the server will be about the same for transactional throughput and response times. Lastly, the same set of measurements is taken on the consolidation target server and results compared.

* including storage and peripheral devices

• As demonstrated in the table, up to 30 older x86 servers can be consolidated to a single SPARC T3-2 server in a secure virtualized environment using Oracle Solaris and Oracle Solaris Containers.
• To conduct this test, 30 Oracle Solaris Containers were configured on the SPARC T3-2 server using the same parameters and workload level as were recorded on each individual x86 server, thus assuring about the same transactional throughput and response times. Since all x86 servers were assumed to be the same, the measurements were repeated on the same server several times to compensate for run-to-run variability.
• The total transaction throughput for the SPARC T3-2 server was within 5 percent of the throughput achieved by 30 x86 servers (4,600 tpm each) that were running at 10 percent CPU utilization.
• The average transaction response time on the SPARC T3-2 server was only slightly higher than the average transaction response time measured on x86 server.
• Despite maintaining about the same transactional throughput and response times, the SPARC T3-2 server had 20 percent of compute capacity available providing for future load increases.
• During the benchmark, the 3U SPARC T3-2 server running at 80 percent CPU utilization consumed 750W. Oracle's Sun Storage F5100 Flash array consumed 350W and occupied one RU. Oracle's Sun Fire X4270 server, used as a COMSTAR target, consumed 300W and occupied 2RU. Oracle's complete solution used only 6RU of space and drew 1400W of power.
• During the benchmark, one two-socket/2RU x86 system running at 10 percent CPU utilization consumed 320W. The complete 30 x86 server configuration consumed a total of 9,600W and occupied 60RU of rack space.
• This dramatic 30:1 consolidation demonstrates a 10x physical footprint reduction, 6.7x better power consumption, and reduced management costs while leaving customers with 20 percent room for additional growth on the SPARC T3-2 server.

Oracle and Java are registered trademarks of Oracle Corporation and/or its affiliates