Mesos: A Platform for Fine-Grained Resource Sharing in the Data Center
Benjamin Hindman, Andrew Konwinski, Matei Zaharia, Ali Ghodsi1, Anthony D. Joseph, Scott Shenker, Ion Stoica and Randy H. Katz
Mesos is a platform for sharing commodity clusters between multiple diverse cluster computing frameworks, such as Hadoop and MPI. Sharing improves cluster utilization and avoids per-framework data replication. Mesos shares resources in a fine-grained manner, allowing frameworks to achieve data locality by taking turns reading data stored on each machine. To support the sophisticated schedulers of today's frameworks, Mesos introduces a distributed two-level scheduling mechanism called resource offers. Mesos decides how many resources to offer each framework, while frameworks decide which resources to accept and which computations to run on them. Our experimental results show that Mesos can achieve near-optimal locality when sharing the cluster among diverse frameworks, can scale up to 50,000 nodes, and is resilient to node failures.
- B. Hindman, A. Konwinski, M. Zaharia and I. Stoica, A Common Substrate for Cluster Computing, HotCloud 2009, June 2009. Available at http://www.usenix.org/events/hotcloud09/tech/full_papers/hindman.pdf
- B. Hindman, A. Konwinski, M. Zaharia, A. Ghodsi, A. D. Joseph, S. Shenker, and I. Stoica, "Nexus: A Common Substrate for Cluster Computing," EECS Department, University of California, Berkeley, Tech. Rep. UCB/EECS-2009-158, Nov. 2009.
- B. Hindman, A. Konwinski, M. Zaharia, A. Ghodsi, A. D. Joseph, R. H. Katz, S. Shenker, and I. Stoica, "Mesos: A Platform for Fine-Grained Resource Sharing in the Data Center," EECS Department, University of California, Berkeley, Tech. Rep. UCB/EECS-2010-87, May 2010.
More information: http://mesos.berkeley.edu