| 摘要 |
<p>System and method for optimizing performance of a distributed computing framework is disclosed. A set of nodes in a distributed computing framework is identified. Subsequently, predefined resources allotted to a node are extracted. A plurality of sub-optimizers is assigned to observe the predefined resources corresponding to the set of nodes. A sub-optimizer submits a linear constraint, to a parent optimizer, corresponding to the predefined resources allotted to the node to execute the job by the node. Based on the predefined resources allotted to the set of nodes, the job is executed using the set of nodes. Based on the execution, performance of the set of nodes executing the job with the resources allotted is monitored. Subsequently, values of the resources allotted to the set of nodes are re-arranged based upon the performance monitored and the linear constraints associated with the set of nodes, to optimize the performance of the distributed computing framework.</p> |
