Driven by ever increasing performance demands of compute-intensive applications, supercomputing systems comprise more and more nodes. This growth is a significant burden for fast group communication primitives and also makes those systems more susceptible to failures of individual nodes. In this paper we present a two-phase fault-tolerant scheme for group communication. Using broadcast as an example, we provide a full-spectrum discussion of our approach — from a formal analysis to LogP-based simulations to a message-passing-based implementation running on a large cluster. Ultimately, we are able to reduce the complex problem of reliable and fault-tolerant collective group communication to a graph theoretical renumbering problem. Both, simulations and measurements, show our solution to achieve a latency reduction of 50% with up to six times fewer messages sent in comparison to existing schemes.
Tue 19 FebDisplayed time zone: Guadalajara, Mexico City, Monterrey change
15:45 - 16:35 | Session 8: HPCMain Conference at Salon 12/13 Chair(s): I-Ting Angelina Lee Washington University in St. Louis | ||
15:45 25mTalk | Corrected Trees for Reliable Group Communication Main Conference Martin Küttler TU Dresden, Maksym Planeta TU Dresden, Germany, Jan Bierbaum TU Dresden, Carsten Weinhold TU Dresden, Hermann Härtig TU Dresden, Amnon Barak The Hebrew University of Jerusalem, Torsten Hoefler ETH Zurich DOI | ||
16:10 25mTalk | Adaptive Sparse Tiling for Sparse Matrix Multiplication Main Conference Changwan Hong , Aravind Sukumaran-Rajam Ohio State University, USA, Israt Nisa , Kunal Singh The Ohio State University, P. Sadayappan Ohio State University DOI | ||