The folks over at the Transaction Processing Performance Council (TPC) have been busy. The TPC benchmarks (such as TPC-C, TPC-D and TPC-H) are the industry standard for benchmarking transaction processing systems that touch upon a broad range of our daily lives, from tracking customer orders and optimizing inventory in warehouses, to supporting critical, real-time business decisions. These benchmarks are the standard by which these types of systems have been measured since their initial release in 1992, and they have been a key factor in research, innovation and performance improvements in relational database systems.
The emergence of new data sources and analytics is pushing IT to innovate in how they process transactions, and toward that end, the TPC recently released a new benchmark for big data technologies called TPCx-HS. Continuing a long history of TPC benchmarks, this new series of tests will enable comparisons of Hadoop architectures (across several dimensions, including overall performance and cost) on an equal footing at the level of scale that is appropriate for every unique deployment.
We’re happy to announce that MapR and Cisco are publishing the first industry benchmark results for TPCx-HS, using the combined solution of the MapR Distribution running on the Cisco UCS platform.
The benchmark configuration consists of the MapR Distribution including Apache Hadoop running on top of the Cisco UCS Integrated Infrastructure (Cisco UCS CPA v2) for Big Data. MapR provides several architectural innovations that were key to completing the benchmark, including an optimized shuffle and MapR-FS for fast data and file access.
The hardware consisted of 16 Cisco UCS C240 M3 Servers that were interconnected using two redundant, active-active Cisco UCS 6296 Fabric Interconnects running the MapR Distribution including Apache Hadoop.
Here are the results of the test, available 01/09/15, with links to the full test results published on the TPC site.
- All results
- 5.07 HSph and $121,231.76/HSph @1TB Scale Factor, Result, ES, FDR
- 5.10 HSph and $120,518.63/HSph @3TB Scale Factor, Result, ES, FDR
- 5.77 HSph and $106,524.27/HSph @10TB Scale Factor, Result, ES, FDR
For more information from Cisco about the benchmark, visit this blog post.
Hadoop performance is key at MapR (you may remember us beating the Terasort world record), and together with Cisco, we’re proud to be the first industry solution to publish results for these new industry benchmarks.