Over the last 30 years, industry standard bodies like the TPC and SPEC have developed many standards for performance benchmarking. The motivation behind these standards is to create technically rigorous, vendor-neutral methods of comparison. These standards have enabled buyers to make more informed decisions about their purchases and have given designers and engineers baselines to better understand their systems, ultimately driving innovation and the development of faster, less expensive, and more energy efficient systems.
Looking back, the most influential and widely adopted standards were the SPEC CPU Benchmark Suites at the system level, and at the application level, the TPC-C (industry standard for benchmarking transaction processing systems), and the TPC-D and its successor, TPC-H (industry standards for benchmarking decision support systems). These were the forerunners to hundreds of benchmark results, appearing in publications and research papers, and driving an ever-expanding list of innovations.
Time marches on and technology-driven innovation continues its relentless advance. Let’s take a closer look at benchmark standards from the TPC in recent years. The TPC has kept pace with the technology, developing and releasing appropriate benchmark standards such as the TPCx-HS and TPCx-BB (benchmark standard for Hadoop based big data analytics), TPCx-DS 2.0 (benchmark standard for decision support on relational and non-relational database systems), and the TPCx-IoT (benchmark standard for IoT gateway systems). In line with the increasing use of virtualization in both private and public clouds, the TPC developed a complete end-to-end virtualization benchmark, TPCx-V.
So, what is TPCx-V designed for? It measures the performance of a server running virtualized databases, and models many properties of virtualized servers including: multiple virtual machines (VMs) running at different load levels, online transaction processing workloads, and decision support system workloads. It uses databases of different sizes and load levels, and simulates large fluctuations in the load levels within virtual machines mirroring real-life load elasticity.
I am a big fan of talking about the industry’s best and first-ever. For those who follow the evolution of database technologies and industry standards, I want to highlight some historical data: the first TPC-C1 and TPC-D2 results were published by IBM; the first TPC-H3 was published by Sun; more recently, the first TPCx-HS4 and TPCx-IoT5 were published by Cisco.
Today, it is my great pleasure to jointly announce the industry’s first ever TPCx-V result. The result was produced using an AMD EPYC™ processor in a Dell EMC server running VMWare.
The benchmark configuration consists of one Dell EMC PowerEdge R7415 with one AMD EPYC 7551P processor (32 core/64 threads), 256 GB DDR4 RAM (2400 Mhz) running VMware ESXi 6.5.0 U2 GA. The TPCx-V throughput performance is 541.5 tpsV and price/performance is 57.31 tpsV/$ . The results were audited by a TPC certified auditor. The full disclosure report can be found here.
Standards-based architectures continue to be the platforms of choice in both private and public clouds, and now AMD has brought choice back to the marketplace. AMD EPYC™ processors offer not only an industry standard based architecture, but many innovations for performance, density and security. I encourage you to learn more about AMD EPYC™ processors in virtualized environments and consider AMD for your next datacenter upgrade cycle.
- First TPC-C publication: 54 tpmC, $ 188,562/tpmC, 12/1995, IBM. Fastest as of today: 30,249,688 tpmC, $ 1.01/tpmC, 12/2010, Oracle
- First TPC-D publication: 84 QthD, $ 52,170/QphD, 09/1992, IBM
- First: TPC-H publication 1,280 QthH, $ 816/QphD, @100GB, 09/1999, Sun. Fastest as of today: 11,612,395 QphH, $ 0.37/QphH @100TB, 9/2014, Dell
- First TPCx-HS publication: 5.07 HSph,$ 121,231.76/HSph @1TB, 1/2015, Cisco. Fastest as of today: 23.42 HSph, $ 36,800/HSph @30TB, 10/2015, Cisco
- First TPCx-IoT publication: 142,493.85 IoTps,$ 0.94/ IoTps, 11/2017, Cisco
Raghu Nambiar is Corporate Vice President & CTO, Datacenter Ecosystem & Application Engineering for AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied. GD-5