High Throughput Computing in the Service of Scientific Discovery

Опубликовано 20 ноября 2019, 21:20

For more than three decades we have been engaged in translating the potential of distributed computing into effective High Throughput Computing (HTC) software tools. The widely adopted HTCondor distributed resource and job management system that we have developed has facilitated evaluation of novel HTC frameworks and technologies under real-life workloads on production distributed systems. Embracing an experimental approach to our work led to close collaboration with researchers from a wide range of domains. These collaborations contributed to two Nobel prizes in recent years and were instrumental in advancing HTC technologies and practices. For example, the UW-Madison Center for High Throughput Computing delivered in the last 12 months close to 400M core hours to more than 250 projects. The talk will present the principles that have been guiding our work, review our experience in deploying HTCondor in different research computing settings, outline our short term development efforts and review long term HTC challenges.

See more at microsoft.com/en-us/research/v...