How to measure the performance of a cloud disk?
The following metrics are generally used to describe the performance of a storage device:
IOPS: The number of reads/writes per second, measured in counts. The underlying driver type of the storage device determines the different IOPS.
Throughput: read/written data volume per second, in MB/s.
Latency: Time that has elapsed from sending an I/O operation to receiving an acknowledgement, in seconds.
How do I test the disk performance?
We recommend using FIO for stress testing and verification of cloud disks. For specific operations, please refer to How to measure the performance of a cloud disk.
Does the I/O size of the application read-write affect the IOPS performance?
Indeed. For a given resource, the IOPS you obtain depends on the I/O size of the application's read and write operations. Typically, the IOPS performance of the disk can be fully utilized during small block reads and writes (for example, when the I/O size is 256KB).
Does the I/O size of the application read-write affect the throughput performance?
Indeed. For a given resource, the throughput you obtain depends on the I/O size of the application's read and write operations. Typically, the disk's throughput performance can be fully utilized during large block reads and writes (for instance, when the I/O size is 1MB).
Can multiple disks be logically combined into one disk to get better performance?
Indeed. You can stripe multiple cloud disks mounted to a cloud server, balance the I/O load across multiple disks, enhance I/O parallelism, and thus achieve higher performance beyond a single disk. For more details, please refer to Building LVM Logical Volumes with Multiple Elastic Cloud Disks.