Multi-Raft configuration for efficient write performances

Multi-Raft configuration improves write performances in Ozone by including DataNodes in multiple pipelines.

Ozone uses the Raft protocol for replicating data across the cluster and providing consistent write performances among the DataNodes. Ozone stores data in a number of containers throughout the cluster and each container allocates data blocks on DataNodes. In addition, Ozone creates pipelines, as logic groups, to assemble containers from several DataNodes for redundancy purposes. Each pipeline consists of DataNodes in a Raft group such that there are three DataNodes with one of them being the leader. Through the pipeline, data is written to the leader, which replicates the same to the followers. A pipeline uses RaftLog to spread the write load on disks.

In a single-Raft configuration, a DataNode can join only one pipeline. This can slow down the write performance to the DataNodes and impact the efficiency with which disks are used on various DataNodes. Starting with CDP 7.1.6, a multi-Raft configuration is available on Ozone by default. A multi-Raft configuration is beneficial because it increases the write efficiency by providing for more containers and pipelines without increasing the number of nodes or disks. Further, multi-Raft configuration also leads to more efficient use of DataNodes and disks in spreading the writes throughout the cluster.

Configuration properties for pipeline limits

To configure the pipeline limits for a multi-Raft configuration, you can set up certain properties as advanced configuration snippets using the Ozone Service Advanced Configuration Snippet (Safety Valve) for ozone-conf/ozone-site.xml property.

Property Description Default value
ozone.scm.datanode.pipeline.limit Limit for the number of 3-node pipelines that a DataNode can join. 2

If you want to increase the number of pipelines based on the number of RaftLog disks, then you can set this value to 0.

ozone.scm.pipeline.per.metadata.disk Number of pipelines to create for a Raft-log disk. 2