Cloud-container storage pool data cache

Edit online

IBM Storage Protect 8.1.12 introduced the CLOUDREADCACHE parameter, which can be used while you are defining or updating a cloud-container storage pool to help improve performance when you restore a large amount of data from cloud-container storage pools. When this parameter is enabled, the system can download a copy of cloud-container object data that is being restored, and then stage a copy of that data locally in the cloud cache. As a result, the cloud cache can store both ingested data and read cache data.

With cloud-container storage pools, unique data extents are stored in a deduplicated, and optionally compressed and encrypted, format in cloud-container objects that are within object storage. Each client file that is backed up to the IBM Storage Protect server is composed of a sequence of deduplicated extents that are stored in these cloud-container objects within object storage. Depending on how client data deduplicates with other client data, individual client files might share data extents in these cloud containers with other files or data that is stored in the cloud-container storage pool.

When you enable the CLOUDREADCACHE parameter for a cloud-container storage pool, the IBM Storage Protect server analyzes container data read requests for the storage pool. Read requests are initiated by client restore and retrieve activity. If significant read activity occurs within a short time for a cloud-container object, the server downloads the requested data from object storage, and then stages it as a copy to the cloud cache. The cloud cache is in local disk storage pool directories. The system uses the read cache data that was downloaded to the cloud cache to meet further client read requests for that cloud-container object. By using a cloud cache, you can improve the performance of restore operations in some circumstances because reading data from local disk storage pool directories can be much faster than sending small read requests to object storage over the Ethernet network.

Enabling the CLOUDREADCACHE parameter might be useful in the following circumstances:
  • The IBM Storage Protect server is connected to the object storage system on a network with high bandwidth. For example, the network bandwidth is 5 Gbps, 10 Gbps, or 40 Gbps.
  • The IBM Storage Protect server is connected to the object storage system on a network with high latency. For example, the object storage system might be an off-premises public cloud with 10 milliseconds of round-trip latency.
  • You must restore more than 10 GB of data.
  • You have an extent workload that is predominantly composed of small files, is stored with a low ratio of data deduplication, or both. For example, you have one of the following types of workload:
    • IBM Storage Protect backup-archive client, file server type workloads
    • Oracle database workloads
  • You would like to restore data multiple times to different clients or for different purposes. For example, you might want to recover a virtual machine or database multiple times when you are setting up test environments.
Enabling the CLOUDREADCACHE parameter might not be useful in the following circumstances:
  • The IBM Storage Protect server is connected to the object storage system on a network with low bandwidth. For example, the network bandwidth is 1 Gbps or 100 Mbps bandwidth.
  • You must restore only a small amount of data (less than 10 GB).
  • You have a large deduplicated extent workload and the IBM Storage Protect server is connected to the object storage system on a network with low latency. For example, the object storage system might be on the same local area network (LAN) as the cloud-container storage pool, and the round-trip latency might be less than 10 milliseconds.
Read cache data invalidation:

Read cache data in the cloud cache is not preserved when the IBM Storage Protect server is restarted. Instead, read cache data is invalidated on a server restart and is no longer used. The cloud cache is initialized, and it prepares storage pool directories.

Read cache data in the cloud cache is invalidated in all of the following situations:
  • The CLOUDREADCACHE parameter value is changed from ON to OFF.
  • The CLOUDREADCACHE parameter value is specified as either ON or ONPREFERINGEST and the IBM Storage Protect server is restarted.
  • The CLOUDREADCACHE parameter value is changed from OFF to either ON or ONPREFERINGEST.