Home > Storage > PowerScale (Isilon) > Industry Solutions and Verticals > Electronic Design Automation > Synthetic EDA Workloads for At-Scale Storage Benchmarking > Conclusion
Measuring EDA workload performance is not a simple process. There are hundreds of EDA S/W applications. Historical benchmarks have attempted to standardize EDA workload I/O patterns, looking for maximum IOPs/Sec as a benchmark measure of performance. But these are of limited value because they fail to consider other factors that also affect EDA workload performance, such as network bandwidth, storage cache performance, number of compute node, job scheduler performance, NFS version, and mount option, and so on.
In real world scenarios, all affect EDA workload performance. More importantly, IOPs is not a direct measure of what customers really care about: they care more about turn-around time (TaT). A good EDA benchmark should represent real-world conditions: at scale, just as most environments are on a daily basis. While the best scenario is to benchmark real EDA workload performance of your environment onsite using your actual EDA workloads, this is not always feasible because the required EDA tool licenses may be in use 24x7 on actual workloads.
This benchmark offers an opportunity to use open-source tools to approximate an EDA workload to benchmark EDA tool infrastructure without the need to tie up costly EDA tool licenses. It also offers portability, because open-source software (in this case, an Android build), is itself license free and can be run in any environment. This offers an outstanding opportunity to compare results – not to mention share infrastructure optimization lessons. Furthermore, this benchmark is not just limited to storage. It can also benchmark your entire EDA environment, including compute node, job scheduler, network utilization, and application servers.
Compared to traditional benchmark tools that focus on peak IOPS, this is a far more realistic and scalable benchmark methodology with a more impactful focus on ‘time to finish’.