This topic describes the performance metrics and specifications of various Block Storage devices, including disks, Shared Block Storage devices, and local disks.

Performance metrics

The key metrics for measuring Block Storage performance are IOPS, throughput, and latency. Some Block Storage devices also have requirements on the capacity. For example, enhanced SSDs of different performance levels (PLs) have different capacity ranges.

  • Input/output operations per second (IOPS)

    The IOPS metric measures the number of write/read operations that can be performed each second. The metric indicates the performance of transaction-intensive applications, such as databases. A standard SSD can achieve the committed IOPS performance only when it is attached to an I/O optimized instance. Common IOPS performance metrics that are measured include sequential and random operations. The following table describes these performance metrics.

    Metric Description Data access method
    Total IOPS The total number of I/O operations per second Randomly and sequentially access locations in storage devices
    Random read IOPS The average number of random read I/O operations per second Randomly access locations in storage devices
    Random write IOPS The average number of random write I/O operations per second
    Sequential read IOPS The average number of sequential read I/O operations per second Sequentially access locations in storage devices
    Sequential write IOPS The average number of sequential write I/O operations per second
  • Throughput

    The throughput metric measures the size of data transferred per second. Unit: MB/s. The metric indicates the performance of applications that require a large number of sequential read/write operations, such as Hadoop offline computing applications.

  • Latency
    Latency is the period of time required for a Block Storage device to process an I/O request. Unit: s, ms, or μs. High latency may cause performance deterioration or errors in applications with requirements for low latency, such as databases.
    • For latency-sensitive applications, we recommend that you use enhanced SSDs, standard SSDs, Shared SSD Block Storage, or local SSDs.
    • For applications that require high throughput but do not require low latency, such as Hadoop offline computing applications, we recommend that you use ECS instances attached with local SATA HDDs in the d1 or d1ne instance family.
  • Capacity
    You cannot use capacity as a metric to measure the performance of Block Storage, but the performance of Block Storage devices varies with capacity. The larger the capacity of a Block Storage device, the stronger the data processing capability of the device. Block Storage devices of the same type have the same I/O performance per unit capacity. However, the performance of each disk increases with capacity until the upper limit performance for that single disk is reached.
    Note Block Storage capacity is measured in binary units, such as kibibyte (KiB), mebibyte (MiB), gibibyte (GiB), or tebibyte (TiB). 1 KiB equals 1,024 bytes. 1 MiB equals 1,024 KiB. 1 GiB equals 1,024 MiB. 1 TiB equals 1,024 GiB.

For information about how to test the performance of different types of Block Storage devices, see Performance tests on Block Storage or Test the IOPS performance of an enhanced SSD.

Disk performance

The following table lists the performance and typical scenarios of different types of disks.
Performance category Enhanced SSD Standard SSD Ultra disk Basic disk***
Performance level (PL) PL3 PL2 PL1 None None None
Maximum capacity of a single disk 1,261-32,768 GiB 461-32,768 GiB 20-32,768 GiB 32,768 GiB 32,768 GiB 2,000 GiB
Maximum IOPS 1,000,000 100,000 50,000 25,000* 5,000 Several hundreds
Maximum throughput 4,000 MB/s 750 MB/s 350 MB/s 300 MB/s* 140 MB/s 30−40 MB/s
Formula for calculating IOPS of a single disk** min{1,800 + 50 × Capacity, 1,000,000} min{1,800 + 50 × Capacity, 100,000} min{1,800 + 50 × Capacity, 50,000} min{1,800 + 30 × Capacity, 25,000} min{1,800 + 8 × Capacity, 5,000} None
Formula for calculating throughput of a single disk** min{120 + 0.5 × Capacity, 4,000} MB/s min{120 + 0.5 × Capacity, 750} MB/s min{120 + 0.5 × Capacity, 350} MB/s min{120 + 0.5 × Capacity, 300} MB/s min{100 + 0.15 × Capacity, 140} MB/s None
Data reliability 99.9999999% 99.9999999% 99.9999999% 99.9999999% 99.9999999% 99.9999999%
Single-channel random write access latency 0.2 ms 0.5−2 ms 1−3 ms 5−10 ms
API parameter value cloud_essd cloud_ssd cloud_efficiency cloud
Scenario
  • OLTP databases: relational databases such as MySQL, PostgreSQL, Oracle, and SQL Server databases
  • NoSQL databases: non-relational databases such as MongoDB, HBase, and Cassandra databases
  • Elasticsearch distributed logs: Elasticsearch, Logstash, and Kibana (ELK) log analysis
Small and medium-sized development and test applications that require high data reliability
  • Development and test applications
  • System disks
  • Applications that are not frequently accessed or have low I/O loads
  • Applications that require low costs and random I/O operations

* The performance of standard SSDs varies with the sizes of data blocks. Smaller data blocks result in lower throughput and higher IOPS, as described in the following table.

Data block size (KiB) Maximum IOPS Throughput (MB/s)
4 Approximately 25,000 Approximately 100
16 Approximately 17,200 Approximately 260
32 Approximately 9,600 Approximately 300
64 Approximately 4,800 Approximately 300

** A standard SSD is used as an example to describe how to calculate the performance of a single disk:

  • Maximum IOPS: The baseline is 1,800 IOPS. It increases by 30 IOPS per GiB of storage. The maximum IOPS is 25,000.
  • Maximum throughput: The baseline is 120 MB/s. It increases by 0.5 MB/s per GiB of storage. The maximum throughput is 300 MB/s.

*** Basic disks are the previous generation of disks and are no longer available for purchase.

Shared Block Storage performance

The following table lists the performance and typical scenarios of different types of Shared Block Storage.

Performance category Shared SSD Block Storage Shared Ultra Block Storage
Maximum capacity of a single disk 32,768 GiB 32,768 GiB
Maximum capacity of a single instance 128 TiB 128 TiB
Maximum random read/write IOPS* 30,000 5,000
Maximum sequential read/write throughput* 512 MB/s 160 MB/s
Formula for calculating the IOPS of a single disk** min{1,600 + 40 × Capacity, 30,000} min{1,000 + 6 × Capacity, 5,000}
Formula for calculating the throughput of a single disk** min{100 + 0.5 × Capacity, 512} MB/s min{50 + 0.15 × Capacity, 160} MB/s
Single-channel access latency 0.5−2 ms 1−3 ms
Scenario
  • Oracle RAC
  • Failover cluster
  • High-availability servers
High-availability architecture of servers

* The maximum IOPS and throughput listed in the preceding table are the maximum performance of a bare Shared Block Storage device that is attached to two or more instances at the same time during stress tests.

** A Shared SSD Block Storage device is used as an example to describe how to calculate the performance of a single disk:

  • Maximum IOPS: The baseline is 1,600 IOPS. It increases by 40 IOPS per GiB of storage. The maximum IOPS is 30,000.
  • Maximum throughput: The baseline is 100 MB/s. It increases by 0.5 MB/s per GiB of storage. The maximum throughput is 512 MB/s.

Local disks

For information about the performance of local NVMe SSDs and SATA HDDs, see Local disks.