240 likes | 438 Views
Background. In 2007, the RAD Lab began exploring data analytics to support developers, focusing on the fast-growing Hadoop platformSeveral projects looked at resource scheduling in MapReduce, leading to some code contributions to Hadoop
E N D
1. Mesos: A Platform for Fine-Grained Resource Sharing in Datacenters Benjamin Hindman, Andy Konwinski, Matei Zaharia,Ali Ghodsi, Anthony D. Joseph, Randy Katz,Scott Shenker, Ion Stoica
2. Background In 2007, the RAD Lab began exploring data analytics to support developers, focusing on the fast-growing Hadoop platform
Several projects looked at resource scheduling in MapReduce, leading to some code contributions to Hadoop & Dryad
3. RAD Lab Work onMapReduce Scheduling LATE algorithm for straggler mitigation(M. Zaharia, A. Konwinski, A. Joseph, R. Katz, I. Stoica)
Energy-efficient MapReduce(Y. Chen, L. Keys, A. Ganapathi, R. Katz)
Hadoop Fair Scheduler and delay scheduling(M. Zaharia, S. Shenker, I. Stoica, and collaborators at Facebook and Yahoo! Research)
Progress fairness (G. Lee and R. Katz)
Mantri straggler mitigation (G. Ananthanarayanan,I. Stoica, and collaborators at Microsoft Research)
Network scheduling (C. Reiss and R. Katz)
Scarlett adaptive replication (G. Ananthanarayanan,S. Agarwal, I. Stoica, and collaborators at Microsoft)
4. Motivation for Mesos Datacenters are running an increasingly diverse set of applications, both for analytics and user facing services Programming models: MapReduce, Dryad, Pregel, … Storage systems: HDFS, HBase, MySQL, … Web applications