IBM Cloud Docs
Block Storage capacity and performance

Block Storage capacity and performance

Choosing the optimal Block Storage volume size and performance level for your workloads is important. When you provision Block Storage for VPC, you can specify the size of your volume and the performance level that you require.

Capacity

Block Storage for VPC offers a range of storage capacities to meet your requirements. Based on the storage profile that you chose for your data volume, you can specify 10-16,000 GB of capacity in 1 GB increments. Boot volumes are 100 GB by default. If you provision an instance from a custom image, you can specify a boot volume capacity up to 250 GB.

Block Storage volume profiles

When you provision Block Storage for VPC volumes, you specify a volume profile that best meets your storage requirements. Three predefined tiered profiles are available, or you can choose a custom profile. Profiles in the tiered family provide pre-defined IOPS/GB performance for volumes up to 16,000 GB capacity. A Custom profile defines ranges of volume capacity and IOPS that you can select. These profiles are backed by solid-state drives (SSDs).

How volume bandwidth is allocated

Bandwidth that is available to the VSI is split between attached Block Storage volumes and networking. The initial volume and network bandwidth allocation depends on the instance profile. The allocation of the instance's total bandwidth can be adjusted, balancing between network bandwidth and volume bandwidth. If you do not specify the initial volume and network bandwidth allocation, then 25% of total instance bandwidth is allocated to volume bandwidth and 75% is allocated to network bandwidth. For more information, see Bandwidth allocation for instance profiles.

The provisioned volume bandwidth is the highest potential bandwidth that can be allocated to the volume when it is attached to an instance. In cases where the total maximum bandwidth of attached volumes exceeds the amount that is available on the instance, the bandwidth for each volume attachment is set proportionally. The bandwidth is allocated based on the corresponding volume's maximum bandwidth. For more information, see Bandwidth allocation for Block Storage volumes.

How I/O size affects performance

The IOPS metric shows how many read and/or write operations a storage device can perform per second. The IOPS value of a volume is based on a 16 KB block size with a 50-50 read/write random workload for all the volume profiles. Each 16 KB of data read/written counts as one read/write operation; a single write of less than 16 KB counts as a single write operation.

The provisioned volume bandwidth is determined by the number of IOPS multiplied by the throughput multiplier. The throughput multiplier is 16 KB for 3 IOPS/GB or 5 IOPS/GB tiered profiles, or 256 KB for 10 IOPS/GB tiered or custom volume profiles. The higher the IOPS that you specify, the higher the throughput. Maximum throughput is 1024 MBps (8192 Mbps). For more information about the maximum throughput values, see Block storage profile families.

Your application's I/O size can be different than the throughput multiplier of the storage volume profile and it directly impacts how well your workload performs. If the application I/O size is smaller than the throughput multiplier, the IOPS limit is reached before the throughput limit. Conversely, if the application I/O size is larger, the throughput limit is reached before the IOPS limit.

The following table provides some examples of how application I/O size and provisioned IOPS affect the throughput, which is calculated as average application I/O size x IOPS = Throughput in MBps.

Examples of how application I/O size and IOPS affect the throughput
Average I/O Size (KB) IOPS Throughput (MBps)
4 (typical for Linux®) 1,000 4
8 (typical for Oracle) 1,000 8
16 1,000 16
32 (typical for SQL Server) 500 16
64 250 16
128 128 16

In these examples, your performance caps are 1000 IOPS or 16 MBps throughput. You can achieve maximum IOPS when you use smaller I/O sizes, but throughput is less than what the volume can handle. The following example shows how throughput decreases for smaller average I/O sizes, when max IOPS is maintained.

  • 16 KB * 6000 IOPS == ~94 MBps
  • 8 KB * 6000 IOPS == ~47 MBps
  • 4 KB * 6000 IOPS == ~23 MBps