Checklist for storage pools on DISK or FILE

Use the checklist to review how your disk storage pools are set up. This checklist includes tips for storage pools that use DISK or FILE device classes.

Question Tasks, characteristics, options, or settings More information
Can the storage pool LUNs sustain throughput rates for 256 KB sequential reads and writes to adequately handle the workload within the time constraints? When you are planning for peak loads, consider all the data that you want the server to read or write to the disk storage pools simultaneously. For example, consider the peak flow of data from client backup operations and server data-movement operations such as migration that run at the same time.

The IBM Spectrum Protect server reads and writes to storage pools predominantly in 256 KB blocks.

If the disk system includes the capability, configure the disk system for optimal performance with sequential read/write operations rather than random read/write operations.

For more information, see Analyzing the basic performance of disk systems.
Did you allocate enough storage space for the database? For a rough estimate, the following database size guidelines are based on the small, medium, and large blueprint systems to allow for database growth:
  • Small system: At least 1 TB
  • Medium system: At least 2 TB
  • Large system: At least 4 TB
Tip: You might need more memory based on the amount of data that must be protected, the number of files that are stored, and whether you use data deduplication. With data deduplication, the load on the database becomes greater because there are frequent queries to the database to determine what deduplicated extents are on the server.

For a rough estimate, plan for 100 GB of database storage for every 50 TB of data that is to be protected in deduplicated storage pools. Protected data is the amount of data before data deduplication, including all versions of objects stored.

If you have several hundred TB of protected data, or if you are backing up multiple TBs of data daily, the starting size for the database must be at least 1 TB. Use the IBM Spectrum Protect to size the database for your system.

The optimal IBM Spectrum Protect environment is set up by using the IBM Spectrum Protect Blueprints.

For information about the minimum amount of memory you must allocate on the server to complete operations, based on the database size, see Memory requirements

Is the disk configured to use read and write cache? Use more cache for better performance.  
Do you need to backup the IBM Spectrum Protect database to cloud object storage?

You can back up a database to, and restore a database from, cloud object storage for disaster recovery purposes.

You can tune object storage endpoints, IBM® Cloud Object Storage Accessers, network bandwidth, and data streams to ensure that database backup operations run efficiently.

Tuning database backups to cloud object storage.
For storage pools that use FILE device classes, have you determined a good size to use for the storage pool volumes? Review the information in Optimal number and size of volumes for storage pools that use disk. If you do not have the information to estimate a size for FILE device class volumes, start with volumes that are 50 GB. Typically, problems arise more frequently when the volumes are too small. Few problems are reported when volumes are larger than needed. When you determine the volume size to use, as a precaution choose a size that might be larger than necessary.
For storage pools that use FILE device classes, are you using preallocated volumes? Scratch volumes can cause file fragmentation.

To ensure that a storage pool does not run out of volumes, set the MAXSCRATCH parameter to a value greater than zero.

Use the DEFINE VOLUME server command to preallocate volumes in the storage pool.

Use the DEFINE STGPOOL or UPDATE STGPOOL server command to set the MAXSCRATCH parameter.

For storage pools that use FILE device classes, did you compare the maximum number of client sessions to the number of volumes that are defined? Always maintain enough usable volumes in the storage pools to allow for the expected peak number of client sessions that run at one time. The volumes might be scratch volumes, empty volumes, or partly filled volumes. For storage pools that use FILE device classes, only one session or process can write to a volume at the same time.
For storage pools that use FILE device classes, have you set the MOUNTLIMIT parameter of the device class to a value that is high enough to account for the number of volumes that might be mounted in parallel? For storage pools that use data deduplication, the MOUNTLIMIT parameter is typically in the range of 500 - 1000.
Set the value for MOUNTLIMIT to the maximum number of mount points that are needed for all active sessions. Consider parameters that affect the maximum number of mount points that are needed:
  • The MAXSESSIONS server option, which is the maximum number of IBM Spectrum Protect sessions that can run concurrently.
  • The MAXNUMMP parameter, which sets the maximum number of mount points that each client node can use.
For example, if the maximum number of client node backup sessions is typically 100 and each of the nodes has MAXNUMMP=2, multiply 100 nodes by the 2 mount points for each node to get the value of 200 for the MOUNTLIMIT parameter.
Use the REGISTER NODE or UPDATE NODE server command to set the MAXNUMMP parameter for client nodes.
For storage pools that use DISK device classes, have you determined how many storage pool volumes to put on each file system? How you configure the storage for a storage pool that uses a DISK device class depends on whether you are using RAID for the disk system.

If you are not using RAID, then configure one file system per physical disk, and define one storage pool volume for each file system.

If you are using RAID 5 with n + 1 volumes, configure the storage in one of the following ways:
  • Configure n file systems on the LUN and define one storage pool volume per file system.
  • Configure one file system and n storage pool volumes for the LUN.
For an example layout that follows this guideline, see Figure 1.
Did you create your storage pools to distribute I/O across multiple file systems? Ensure that each file system is on a different LUN on the disk system.

Typically, having 10 - 30 file systems is a good goal, but ensure that the file systems are no smaller than approximately 250 GB.

For details, see the following topics:
Have you scheduled audit operations to identify corrupted files in storage pools? To schedule audit operations, use the DEFINE STGRULE command and specify the ACTIONTYPE=AUDIT parameter.

To help optimize audit operations and ensure that they run continuously, do not specify the DELAY parameter.