Apparatus and method for managing resources in cluster computing environment

A resource manager node, comprising:

• a resource management unit for performing an operation necessary for resource allocation when a resource allocation request for performance of a specific task on a cluster system including heterogeneous resources is received;
• a resource policy management unit for managing a resource characteristic-based resource allocation policy and a quantitative resource allocation policy, determining a resource allocation policy being the resource characteristic-based resource allocation policy or the quantitative resource allocation policy based on a characteristic of the task, and generating resource allocation information including information about one or more allocated node and resources based on the determined resource allocation policy;
• a shared resource capability management unit for managing each topology of a plurality of nodes connected to the resource manager node, information about capabilities of resources of each of the nodes, and resource association information of each of the nodes;
• a shared resource status monitoring unit for monitoring and managing information about status of each of the plurality of nodes and information about use of allocated resources;
• and
• a shared resource allocation unit for sending a resource allocation request according to the resource allocation information to at least one of the plurality of the nodes based on the resource allocation information, the topology information, and the resource capability information,
• wherein the resource policy management unit, if the resource allocation request is a multi-node request and the resource manager node corresponds to a communication-based parallel model comprising a Message Passing Interface (MPI) to determine communication costs between nodes that can be allocated to a parallel program in the cluster system, generates a first node list including nodes having a least communication cost based on the topology information of the shared resource capability management unit, generates a second node list including nodes belonging to the nodes included in the first node list and meeting a resource association condition by referring to the resource association information of the shared resource capability management unit, and generates the resource allocation information so that resources are provided by a node group having lowest node load within the second node list by referring to the monitoring information,
• wherein the resource policy management unit, if the resource allocation request is a multi-node request and the resource manager node does not correspond to the communication-based parallel model, generates a third node list including nodes meeting a resource association condition by referring to the resource association information of the shared resource capability management unit, and generates the resource allocation information so that resources are provided by a node group having lowest node load within the third node list by referring to the monitoring information,
• wherein the resource policy management unit, if the resource allocation request is a single node request and the determined resource allocation policy is not the quantitative resource allocation policy, determines the characteristic of the task, generates a fourth node list including nodes having resources matching the determined characteristic based on the resource capability information of the shared resource capability management unit, and generates the resource allocation information so that resources are provided by a node having lowest node load within the fourth node list by referring to the monitoring information, and
• wherein the resource policy management unit, if the resource allocation request is a single node request and the determined resource allocation policy corresponds to the quantitative resource allocation policy, generates a fifth node list including nodes belonging to all nodes connected to the resource manager node and being able to allocate resources in response to the resource allocation request using the shared resource capability management unit, generates a sixth node list including nodes belonging to the nodes included in the fifth node list and meeting a resource association condition by referring to the resource association information of the shared resource capability management unit, and generates the resource allocation information so that resources are provided by a node having lowest node load within the sixth node list by referring to the monitoring information.