OLVM Backup Performance Tuning (bp.conf)
NetBackup transfers OLVM disk data using an asynchronous, pipelined mechanism. You can tune the transfer behavior by adding the following keys to the bp.conf file on the backup host. Changes take effect on the next backup job. No service restart is required.
Table:
bp.conf Key | Default Value | Description |
|---|---|---|
| 1 | Enables (1) or disables (0) the asynchronous pipelined disk transfer. When disabled, NetBackup falls back to synchronous I/O. NetBackup also falls back automatically if the async pipeline fails to initialize. |
| 16 | Number of parallel read threads used during backup. Increase for high-throughput storage; decrease if the Backup Host or OLVM engine is resource-constrained. |
| 8 | Number of parallel write threads used during restore. Adjust in proportion to OLVM_ASYNC_READ_THREADS if changing restore performance. |
| 16 | Size of each transfer buffer, in megabytes. Shared between backup and restore pipelines. Larger buffers can improve throughput on high-latency networks. Smaller buffers reduce memory usage. |
| 16 | Number of buffers in the pool per job. Shared between backup and restore pipelines. The working set per job is approximately OLVM_ASYNC_BUFFER_SIZE_MB × OLVM_ASYNC_BUFFER_COUNT (default: 256 MB). |
| 172800 (48 hours) | Timeout, in seconds, for inactive OLVM image transfer sessions. Increase on slow or high-latency networks to prevent premature session termination. |
| 3 | Number of retry attempts per chunk before the async transfer marks the chunk as failed and triggers a job error. Set to 0 to disable retries. |
| (secure) | Set to No to disable TLS certificate verification. |
Note:
The buffer size (OLVM_ASYNC_BUFFER_SIZE_MB) and buffer count (OLVM_ASYNC_BUFFER_COUNT) are shared between the backup and restore pipelines. Changing these values affects both operations. The thread counts (OLVM_ASYNC_READ_THREADS, OLVM_ASYNC_WRITE_THREADS) are independent for backup and restore.