Monitoring write metrics for remote storage
Native Cloud Object Storage includes added or changed monitoring elements for write storage to expose the additional layers in the remote storage hierarchy.
Write storage hierarchy
For Native Cloud Object Storage, the write storage hierarchy has three distinct layers:
- Persistent storage: also referred to as reliable, durable file storage layer. Persistent storage is used by remote storage groups for their Write-Ahead-Log (WAL) to ensure the efficient persistence of the small write operations.
- Local caching tier: supported by fast NVMe drives, maintains a larger working set than what can be maintained in the in-memory cache in the buffer pool. The performance of the local caching lowers the cost of accessing the remote storage.
- Cloud object storage: the main persistence target for the data within the remote storage groups.
Monitoring writes
Note: The following table gives examples of monitoring metrics used for monitoring writes.
Metric name | Description of metric |
---|---|
LOCAL_TIER_WRITE_TIME | The elapsed time in milliseconds spent on writes to the local storage tier. |
CACHING_TIER_WRITE_TIME | The elapsed time in milliseconds spent on writes through the caching tier, excluding the remote storage. |
REMOTE_STORAGE_TIER_WRITE_TIME | The elapsed time in milliseconds spent on writes to the remote storage, excluding the caching tier. |
POOL_WRITE_TIME | The elapsed time in milliseconds spent on writes for writes to either local storage tier, caching tier, or remote storage. |
DIRECT_TIME | The elapsed time in milliseconds spent on direct writes. |
COMPACTION_WRITE_TIME | The elapsed time in milliseconds spent on writes by the background compaction process. |