Arkouda Nightly Performance Graphs
This page contains performance graphs for Arkouda on a single node server, 16 nodes of an InfiniBand cluster, and 16 nodes of a Cray XC supercomputer.
Arkouda is a Python package that implements a subset of the NumPy and Pandas interfaces. It is designed to support exploratory data analysis (EDA) for data scientists at HPC scale. Arkouda enables analyzing multi-terabyte data sets at interactive rates in order to test hypotheses and react to the results within the human thought loop.
Arkouda is designed to run and scale on a wide variety of platforms. These graphs track the performance of some key benchmarks using a nightly snapshot of Arkouda and the latest stable Chapel release. Results comparing against a nightly snapshot of Chapel can also be enabled.
Configurations:
chapcs comm=none benchmark results on a single node with 24 cores and 256 GB RAM
16-node-cs-hdr comm=gasnet-ibv benchmark results on 16 nodes of a Cray CS cluster with a 200 Gb/s HDR InfiniBand network. 128 cores and 256 GB RAM per node
16-node-xc comm=ugni benchmark results on 16 nodes of a Cray XC with an Aries network. 48 cores and 192 GB RAM per node