Veritas NetBackup™ Appliance Capacity Planning and Performance Tuning Guide
- About this Guide
- Section I. Capacity planning
- Analyzing your backup requirements
- Designing your backup system
- Section II. Best Practices
- Section III. Performance tuning
- Section IV. Quick reference to Capacity planning and Performance tuning
RAID Controller operations
RAID controller plays a very important role in the performance of the appliance. There is a number of operations performed by the RAID controller that can severely influence performance disk and they are analyzed in this section. Table: RAID Controller operations lists the RAID processes performed by the RAID controller and the effect it has on the performance of the appliance:
Table: RAID Controller operations
RAID process | Description | Guidelines |
---|---|---|
RAID Rebuild Process | Disk storage on the NetBackup 52xx appliances is protected using RAID6 wherein a hot spare disk is made available on each shelf. In case of disk failure, the hot spare disk will be used to replace the data disk and a RAID rebuild process will be initialized. | RAID rebuild is very demanding on the disk I/O and RAID controller resources. Since we are using 2TB and 3TB hard disks in appliances, there is a significant amount of data to move and a lot of RAID calculations to perform. This resource intense operation causes a major slowdown of appliance performance, affecting most appliance operations especially tape out, restore or AdvancedDisk operations, since they rely heavily on disk I/O for performance. Placing a load like Backup, Restore, Tape Out, etc. on the appliance while performing RAID rebuild will slow down the rebuild operation, and significantly extend the amount of time needed for the RAID rebuild to finish. |
BBU Learn Cycle Note: This process applies only to the NetBackup 5220 Backup Appliance. | NetBackup 5220 Backup Appliance has Li-ion Battery Backup Unit (BBU) used to protect the RAID controller write cache in the case of complete loss of power. If for any reason BBU is not operational, the RAID controller will switch off write cache (making it write-through) and this in turn will significantly slow down appliance disk I/O and affect performance of all operations on the appliance. | BBU learn cycle usually takes around 3-4 hours to complete the full cycle of discharge and recharge. Appliance status displays that the battery is going through the learn cycle. Although this performance drop is unavoidable, it is a rare event since it happens once in 30 days and high performance operations should be planned around it. If RAID is going through the rebuild process it is recommended to reschedule BBU Learn cycle to avoid a slowdown in RAID rebuild performance. This can be done by the customer support engagement. It is possible to set amount of time until the next BBU Learn Cycle by contacting Support. The Support team can help you to set the number of seconds before next BBU Learn Cycle starts, counting from the last appliance restart. Although this method is not very exact, it allows for some kind of predictability and makes it possible to plan for high performance operations on appliance around this time. For more information on setting BBU Learn Cycle clock, refer to the tech note TECH204348. |
RAID consistency check | This is a regular consistency check operation performed on all LUN's on the RAID controller every 30 days. This check helps confirm that the integrity of RAID group is intact. | It can influence the performance of the disk I/O operations but is not as significant as BBU Learn cycle. The default utilization of resources will be limited to 30% when the consistency check operation takes place. It is recommended to avoid overlapping high performance loads with consistency check. For example, if the BBU Learn cycle on the NetBackup 5220 Backup Appliance if overlapped with consistency check, it will slow down consistency check performance significantly. It is recommended to schedule these operations separately. |
RAID Patrol Read | RAID Patrol Read reads check physical hard disks for media errors. This operation is scheduled to run every seven days and is run simultaneously on a small number of physical disks at the same time. | Patrol reads do not affect performance significantly since they are limited to the few disks at a time, but they still affect disk I/O on the checked disks. It is a best practice to make sure that patrol reads do not overlap other maintenance or recovery operations. |