Help Center/ MapReduce Service/ User Guide/ Managing Clusters/ Managing Static Service Resources in an MRS Cluster/ Overview of Static Service Resources
Updated on 2024-09-23 GMT+08:00
View PDF
Share

Overview of Static Service Resources

Introduction

A cluster allocates static service resources to services Flume, HBase, HDFS, and YARN. The total volume of computing resources allocated to each service is fixed, and they are static. A tenant can exclusively use or share a service to obtain the resources required for running this service.

Static Service Pool

Static service pools are used to specify service resource configurations.

Static service pools centrally manage resources that can be used by each service.

  • Limits the total number of resources that can be used by each service. Specifically, the total number of CPU, I/O, and memory resources can be configured on the nodes where services Flume, HBase, HDFSIoTDB, Kafka (Kafka supports static service pools only in MRS 3.2.0 or later), and YARN are deployed.
  • Isolates the resources of services in a cluster from those of other services. In this way, the load of one service has very limited impact on other services.

Scheduling Mechanism

The time-based dynamic resource scheduling mechanism enables different volumes of static resources to be configured for services at different time, optimizing service running environments and improving the cluster efficiency.

In a complex cluster environment, multiple services share resources in the cluster, but the resource service period of each service may be different.

The following use a bank customer as an example:

  • The HBase query service is heavy in the daytime.
  • The query service is light, but the Hive analysis service is heavy at night.

If fixed resources are allocated to each service, the following problems may occur:

  • The query service cannot obtain sufficient resources while the resources for the analysis service are idle in the daytime.
  • The analysis service cannot obtain sufficient resources while the resources for the query service are idle at night.

As a result, the cluster resource utilization is low and the service capability is weak. Resolve the problem in the following ways:

  • Sufficient resources need to be configured for HBase in the daytime.
  • Sufficient resources need to be configured for Hive at night.

The time-based dynamic scheduling mechanism can efficiently utilize resources and run tasks.