EEMBC Tackles Performance Benchmarks for Cloud and Big Data Workloads with New Working Group
Benchmarks to measure latency and throughput of scale-out servers
EL DORADO HILLS, Calif. — Oct. 15, 2014 — The Embedded Microprocessor Benchmark Consortium (EEMBC) today announced the formation of its new Cloud and Big-Data Server working group. With a charter to build a standardized, industry-endorsed suite of performance and efficiency benchmarks that characterize system-on-chip (SoC) performance for modern cloud and big-data related workloads (scale-out computing), the benchmarks could ultimately lead to the acceleration of a wide variety of real-world applications, including those used in unstructured data stores, web caching, web serving, elastic search, media streaming, data analytics, distributed cloud storage, and graph analytics.
These benchmarks aim to address the needs of ODMS, OEMS, and cloud and big-data users who deploy their applications in large distributed computer centers made up of clusters of servers. To date, the industry has lacked a reliable, repeatable, portable, and architecture-neutral method to evaluate the latency and throughput of SoCs and associated servers for real-world cloud and big-data applications running at data centers.
These applications typically run on expensive clusters of machines with 10s to 1000s of nodes. These may run as a virtual data center inside a physical datacenter (the cloud). This EEMBC benchmark will run on small, relatively inexpensive systems and be able to accurately extrapolate the results to larger deployments, an important differentiator from typical workload benchmarks that require full-blown systems.
“The beauty of the EEMBC benchmark will be its simplicity and accessibility in meeting our vision focused on cloud and big-data analysis and application scalability,” said Markus Levy, EEMBC’s president.“ Aligned with EEMBC’s mission to focus on benchmarks that mimic real-world customer applications, this new benchmark targets the scale-out servers used for specific cloud applications.”
“In our analysis of processors and systems targeting the cloud and big data applications, we found a lack of benchmarks that are representative of real-world cloud workloads,” said Linley Gwennap, principal analyst of The Linley Group. “We are excited to see that EEMBC is tackling this problem with their new Cloud and Big Data working group initiative and encourage cloud system hosts and cloud system developers to get involved in this effort to help EEMBC deliver fair and equitable benchmarks for this growing market segment.”
The Cloud and Big-Data Server working group, led by Narayan Iyengar, lead software engineer at Cavium, Inc., is comprised of EEMBC members with strong connections to the server market. These include Cavium, Imagination Technologies, Intel, and others.
“Cloud and big-data are clearly the wave of the future and the industry needs a broad range of tools, including real-world benchmarks, to allow users to more accurately predict performance of their applications in scale-out deployments,” said Syed Ali, CEO of Cavium, Inc. “EEMBC’s 17 years of experience in developing real-world benchmarks for the industry will ensure that this working group creates a credible benchmark from which our customers and our customer’s customers will derive great value.”
On Oct. 22, representatives from EEMBC will be exhibiting and discussing this new cloud and big data benchmark at the Linley Processor Conference in Santa Clara, Calif. On Oct. 23, attendees will also learn more about this benchmark in Gopal Hedge’s presentation entitled, “Workload-Optimized Processor Architectures for Next-Generation Data Centers and Cloud.”
EEMBC encourages vendors and manufacturers to join the consortium’s working group to contribute to the definition and development of our next generation Cloud and Big Data Server benchmark. To join the working group, contact Markus Levy for details.