130 likes | 278 Views
An Evaluation of Current High-Performance Networks. Christian Bell, Dan Bonachea, Yannick Cote, Jason Duell, Paul Hargrove, Parry Husbands, Costin Iancu, Michael Welcome, Kathy Yelick Lawrence Berkeley National Lab & U.C. Berkeley. http://upc.lbl.gov. Motivation.
E N D
An Evaluation of Current High-Performance Networks Christian Bell, Dan Bonachea, Yannick Cote, Jason Duell, Paul Hargrove, Parry Husbands, Costin Iancu, Michael Welcome, Kathy Yelick Lawrence Berkeley National Lab & U.C. Berkeley http://upc.lbl.gov
Motivation • Benchmark a variety of current high-speed Networks • Measure Latency and Software Overhead, not just Bandwidth • One-sided communication provides advantages vs. 2-sided MPI? • Global Address Space(GAS) Languages • UPC, Titanium (Java), Co-Array Fortran • Small message performance (8 bytes) • Support sparse/irregular/adaptive programs • Programming model: incremental optimization • Overlapping messages can hide the latency
LogGP: no overlap P0 P0 osend osend L orecv orecv P1 P1 Modified LogGP Model • Observed: overheads can overlap: L can be negative EEL: end to end latency (instead of transport latency L) g: minimum time between small message sends G: additional gap per byte for larger messages
Microbenchmarks Ping-pong test: measures EEL (end-to-end latency) Flood test: measures gap (g/G) CPU overlap test: measures software overheads P0 P0 P0 osend osend osend gap gap gap cpu cpu CPU Test 1 CPU Test 2 Flood Test
Gap varies with msg clustering Clustering messages can both use idle cycles, and reduce the number of idle cycles that need to be filled
Potential for CPU overlap during clustered message sends Hardware support for 1-way communication provides more opportunity for computational overlap
“Large” Messages Factor of 6 between minimum sizes needed for “large” message (large = bandwidth dominates fixed message cost)
Small message performance over time Software send overhead for 8-byte messages over time. Not improving much over time (even in absolute terms)
Conclusion • Latency and software overhead of messages varies widely among today’s HPC networks • Affects ability to effectively mask communication latency, with large effect on GAS language viability • especially software overhead--latency can be hidden • These parameters have historically been overlooked in benchmarks and vendor evaluations • Hopefully this will change • Recent discussions with vendors promising • Incorporation into standard benchmarks would be nice… http://upc.lbl.gov