Sizing a cloud cache to optimize backup operations
Before you back up data to a cloud-container storage pool, size the cloud cache. By correctly sizing the cache, you can help to improve the throughput of backup operations and reduce the risk of backup failures.
Before you begin
Tip: If you plan to use cloud tiering to move data from one or more
directory-container storage pools to a cloud-container storage pool, you do not have to size the
cloud cache. Size the cloud cache only for cloud-container storage pools that are the target of
backup operations.
Review the following figure to learn about disk technologies that you can
use to size the cloud cache. The optimum size of the cloud cache depends on the speed of the network
connection and the throughput capability of the object storage system. The figure shows a range of 1
GB to 10 GB Ethernet network connection. For systems with slower network connections and less
throughput capability to cloud object storage, you can use slower and larger disk technologies for
the cloud cache. To optimize throughput, use fast disk technology that can write and read data at
your network speed simultaneously with a 256 KiB input/output (I/O) size. The following example
storage systems are used:- Fast network connection (more than 10 GB): 1 IBM FlashSystem® Drive (3 TB) or 5 IBM Cloud® BMS 960 GB Solid State Disk (SSD) arrays, for example, RAID 5 arrays
- Fast network connection (8 - 10 GB): 2 IBM® StorWize V5000 SSD arrays, for example, 5 TB RAID 5 arrays or 3 IBM Cloud BMS 960 GB SSDs, for example, RAID 5 arrays
- Medium network connection (4 - 8 GB): 1 IBM StorWize V5000 serial-attached SCSI (SAS) and 2 15,000 RAID 5 volumes
- Slow network connection (1 - 4 GB): 1 IBM StorWize V5000 SAS

About this task
For cloud-container storage pools, the data is held only temporarily on disk. After the data is transferred to the cloud, the data is deleted from the cloud cache.
Client data is ingested to the cloud cache at the same time that other ingested data is transferred to the cloud. To prevent the failure of backup operations, size the cloud cache in terms of I/O operations per second (IOPS) and capacity.
Procedure
To size the cloud cache, use the following guidelines:
- Size the cloud cache twice as large as the largest number of data backups that are ingested at the same time. If twice as large is not possible, size the cloud cache at least as large as the largest number of data backups that occur at once.
- To optimize data ingestion, choose fast disk technology, such as SSD, for the cloud cache that can both write and read data at your network speed simultaneously. If you size the cloud cache too large and on slow disk technology, only one copy of the data backup exists. If the cloud cache fails, you lose the data backup. Slow disk technology can cause the cloud cache to be a bottleneck to the system's ingest capability.
- Use RAID 5, RAID 6, or other disk protection for the cloud cache to avoid data loss.
- Use benchmarking tools to help you to size the cloud cache. The cloud benchmarking tools and example benchmarking tests are available on the wiki page at Cloud Blueprints. Benchmark the IOPS capability of the cloud cache and the throughput capability of the network and object storage. For cloud cache benchmarking, the cloud benchmarking tools use an I/O size of 256 KB. The 256 KB size supports simultaneous read-and-write operations with many threads.
What to do next
- Configure a storage pool directory for each cloud-container storage pool. The storage pool directory will define the size of the disk cache. For instructions, see Optimizing performance for cloud object storage.
- Use a single file system, with striping across multiple volumes. Use a stripe size of 16 KB for the cloud cache and ensure that striping is enabled across all physical volumes. Use a single storage pool directory and file system for the cloud cache to avoid disk hot spots with the overlapped I/O of client data ingestion and transfer to cloud.