Dynamically adaptive, resource aware system and method for scheduling
The following relates generally to computer system efficiency improvements. Broadly, systems and methods are disclosed that improve efficiency in a cluster of nodes by efficient processing of tasks among nodes in the cluster of nodes. Assignment of tasks to compute nodes may be based on learned CPU capabilities and I/O bandwidth capabilities of the compute nodes in the cluster.
1. A system for scheduling jobs, said system comprising:
a plurality of interconnected compute nodes defining a cluster of compute nodes, the cluster including a NameNode and a multitude of DataNodes;
the NameNode including at least one processor programmed to:
learn central processing unit (CPU) capabilities and disk input/output (I/O) bandwidth capabilities of compute nodes in the cluster;
determine an optimal number of containers for each application running on each DataNode of the multitude of DataNodes; and
schedule execution of a plurality of tasks on compute nodes of the cluster based on: (i) the learned CPU capabilities and I/O bandwidth capabilities of compute nodes of the cluster, and (ii) at least one of the determined optimal number of containers;
wherein the at least one processor is further programmed to schedule the execution of the plurality of tasks on compute nodes by:
computing a current under-allocation for an application;
and
comparing the current under-allocation to a determined optimal number of containers;
wherein the at least one processor is further programmed to:
determine the optimal number of containers so as to maximize throughput; and
calculate the throughout based on a normalized number of completed map tasks, wherein the normalization is done based on sizes of the map tasks.
2. The system of claim 1 , wherein the at least one processor is further programmed to schedule the execution of the plurality of tasks on compute nodes of the cluster further based on each of the determined optimal number of containers.
3. The system of claim 1 , wherein the at least one processor is further programmed to determine the optimal number of containers using an argmax equation.
4. The system of claim 1 , wherein the at least one processor is further programmed to schedule the execution of the plurality of tasks on compute nodes by:
not considering task combinations that would oversubscribe memory of a compute node of the cluster.
5. The system of claim 4 , wherein the at least one processor is further programmed to schedule the execution of the plurality of tasks on compute nodes by:
validating that the memory of the compute node of the cluster is not oversubscribed by monitoring memory used by each compute node.
6. A method for scheduling jobs in a cluster of compute nodes including a NameNode and a multitude of DataNodes, said method comprising:
learning central processing unit (CPU) capabilities and disk input/output (I/O) bandwidth capabilities of compute nodes in the cluster;
determining an optimal number of containers for each application running on each DataNode of the multitude of DataNodes; and
scheduling execution of a plurality of tasks on compute nodes of the cluster based on: (i) the learned CPU capabilities and I/O bandwidth capabilities of compute nodes of the cluster, and (ii) at least one of the determined optimal number of containers;
wherein the scheduling further comprises:
computing a current under-allocation for an application; and
comparing the current under-allocation to a determined optimal number of containers;
wherein the method further comprises:
determining the optimal number of containers so as to maximize throughput; and
calculating the throughput based on a normalized number of completed map tasks, wherein the normalization is done based on sizes of the map tasks.
7. The method of claim 6 , wherein the scheduling is further based on each of the determined optimal number of containers.
8. The method of claim 6 , wherein the determining further comprises determining the optimal number of containers using an argmax equation.
9. The method of claim 6 , wherein the scheduling further comprises:
not considering task combinations that would oversubscribe memory of a compute node of the cluster.
10. The method of claim 9 , wherein the scheduling further comprises:
validating that the memory of the compute node of the cluster is not oversubscribed by monitoring memory used by each compute node.