Performance considerations

Understand the performance considerations that affect a storage cluster's performance when adding or removing Ceph OSD nodes.

The following factors typically affect a storage cluster’s performance when adding or removing Ceph OSD nodes:

  • Ceph clients place load on the I/O interface to Ceph; that is, the clients place load on a pool. A pool maps to a CRUSH ruleset. The underlying CRUSH hierarchy allows Ceph to place data across failure domains. If the underlying Ceph OSD node involves a pool that is experiencing high client load, the client load could significantly affect recovery time and reduce performance. Because write operations require data replication for durability, write-intensive client loads in particular can increase the time for the storage cluster to recover.

  • Generally, the capacity you are adding or removing affects the storage cluster’s time to recover. In addition, the storage density of the node you add or remove might also affect recovery times. For example, a node with 36 OSDs typically takes longer to recover than a node with 12 OSDs.

  • When removing nodes, you MUST ensure that you have sufficient spare capacity so that you will not reach full ratio or near full ratio. If the storage cluster reaches full ratio, Ceph will suspend write operations to prevent data loss.

  • A Ceph OSD node maps to at least one Ceph CRUSH hierarchy, and the hierarchy maps to at least one pool. Each pool that uses a CRUSH ruleset experiences a performance impact when Ceph OSD nodes are added or removed.

  • Replication pools tend to use more network bandwidth to replicate deep copies of the data, whereas erasure coded pools tend to use more CPU to calculate k+m coding chunks. The more copies that exist of the data, the longer it takes for the storage cluster to recover. For example, a larger pool or one that has a greater number of k+m chunks will take longer to recover than a replication pool with fewer copies of the same data.

  • Drives, controllers and network interface cards all have throughput characteristics that might impact the recovery time. Generally, nodes with higher throughput characteristics, such as 10 Gbps and SSDs, recover more quickly than nodes with lower throughput characteristics, such as 1 Gbps and SATA drives.