The FlexPod Performance Advantage for SQL Server
2016-03-25 11:22 AM
March / April 2016
Cisco and NetApp have teamed for more than five years developing FlexPod® converged infrastructure solutions—with more than 100 documented designs published so far. These solutions combine NetApp® storage systems, Cisco® Unified Computing System servers, and Cisco Nexus fabric switches into validated architectures. They provide a highly efficient operational model for thousands of workload deployments.
FlexPod has proven to be an ideal solution for performance-intensive workloads, such as databases, virtual desktops, and heavily consolidated virtual server workloads. To illustrate FlexPod performance characteristics in more detail, this article examines two use cases that feature Microsoft® SQL Server.
Twenty-Times Reduction in I/O Latency in HammerDB Tests
NetApp recently performed comparison tests using SQL Server 2014 and the open source HammerDB workload generator to simulate an OLTP environment, driven simultaneously from 10 database servers. These tests validated that a FlexPod configuration equipped with NetApp all-flash storage delivered superior performance and required fewer servers, storage devices, and network connection points than a traditional HDD-based system.
Initially, the SQL Server workload was directed to a legacy storage array containing 144 450GB 15K HDDs. The database load was increased until consistent read latencies of 20ms (the threshold of unacceptable performance) were observed. After this baseline performance was captured, identical database and HammerDB configurations were directed to a NetApp 8080 EX All Flash FAS (AFF) configured with 48 400GB SSDs.
The results were striking: 20-times reduction in I/O latencies, four-times improvement in storage IOPS, and four-times improvement in SQL Server CPU utilization. In addition, the performance headroom remaining on the AFF8080 EX would have enabled it to deliver further performance increases over and above what was observed in the comparison tests.
These tests demonstrate the massive performance advantages of FlexPod systems using all-flash arrays. The tests also demonstrate FlexPod systems’ potential to reduce the costs for associated server gear, network infrastructure, and per-CPU software licensing fees in SQL Server environments.
208% Faster SQL Server Response Time
Scalability Experts, a triple Gold Microsoft Partner, performed SQL Server 2014 load tests using a real-world online gaming workload from Iforium. The testers employed a FlexPod solution consisting of Cisco UCS blade servers, Cisco Nexus unified fabric switches, and a NetApp FAS8020 hybrid storage array configured with 48 600GB SAS HDDs and 1TB of NetApp Flash Cache® intelligent caching.
The testing took place over several weeks as "spins" of virtual slot machines were simulated using the Iforium software platform. During the tests, users were added in increments of 5,000 to 10,000, up to a maximum of 50,000 users. As the graph below indicates, the FlexPod system scaled linearly as more users came onto the platform, with no noticeable impact on response time for up to 30,000 users and 2,000 spins per second. The tests also showed that the FlexPod system could sustain burst rates of up to 50,000 users and nearly 3,000 spins per second, as might occur during a special gaming promotion.
Average spins per second is a central metric for the online gaming business, similar to the way that the number of transactions per second is a critical measure for other types of online retail companies. When Scalability Experts compared its results with SQL Server testing norms and Iforium’s prior workload experience, the company confirmed that the FlexPod configuration, with a relatively modest amount of flash acceleration, provided 208% faster SQL Server transaction response times. The configuration also provided 250% faster page response times, very sequential linear scalability, and a three-times overall performance improvement.
Test Your Own Applications on FlexPod
When optimizing performance, every microsecond counts. Contact your local NetApp representative or partner to request your own Proof of Concept test with an all-flash FlexPod solution.
This NetApp Community is public and open website that is indexed by search engines such as Google. Participation in the NetApp Community is voluntary. All content posted on the NetApp Community is publicly viewable and available. This includes the rich text editor which is not encrypted for https.
- Software files (compressed or uncompressed)
- Files that require an End User License Agreement (EULA)
- Confidential information
- Personal data you do not want publicly available
- Another’s personally identifiable information
- Copyrighted materials without the permission of the copyright owner