CRUM: Checkpoint-Restart Support for CUDA's Unified Memory
| Autor: | Michael J. Sullivan, Rohan Garg, Gene Cooperman, Apoorve Mohan |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2018 |
| Předmět: |
FOS: Computer and information sciences
Unix 020203 distributed computing Computer science Computation 05 social sciences 050301 education 02 engineering and technology Pascal (programming language) Parallel computing Software_PROGRAMMINGTECHNIQUES CUDA Memory management Computer Science - Distributed Parallel and Cluster Computing Virtual memory Scalability 0202 electrical engineering electronic engineering information engineering Stable storage Distributed Parallel and Cluster Computing (cs.DC) 0503 education computer computer.programming_language |
| Zdroj: | CLUSTER |
| Popis: | Unified Virtual Memory (UVM) was recently introduced on recent NVIDIA GPUs. Through software and hardware support, UVM provides a coherent shared memory across the entire heterogeneous node, migrating data as appropriate. The older CUDA programming style is akin to older large-memory UNIX applications which used to directly load and unload memory segments. Newer CUDA programs have started taking advantage of UVM for the same reasons of superior programmability that UNIX applications long ago switched to assuming the presence of virtual memory. Therefore, checkpointing of UVM will become increasingly important, especially as NVIDIA CUDA continues to gain wider popularity: 87 of the top 500 supercomputers in the latest listings are GPU-accelerated, with a current trend of ten additional GPU-based supercomputers each year. A new scalable checkpointing mechanism, CRUM (Checkpoint-Restart for Unified Memory), is demonstrated for hybrid CUDA/MPI computations across multiple computer nodes. CRUM supports a fast, forked checkpointing, which mostly overlaps the CUDA computation with storage of the checkpoint image in stable storage. The runtime overhead of using CRUM is 6% on average, and the time for forked checkpointing is seen to be a factor of up to 40 times less than traditional, synchronous checkpointing. 22 pages; 5 figures; accepted at IEEE Cluster-2018 |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|