Block storage performance indicators and block storage performance specifications _ ECS

The performance and price of different types of block storage vary. You can select the appropriate block storage products according to the actual workload and application needs to meet your storage needs. This article introduces cloud disks and local disks And flexible temporary disk Performance metrics and performance specifications.

explain

For the price and billing information of different types of block storage, see Block storage charging 。
If you want to know the characteristics and application scenarios of different block storage, see Overview of Block Storage 。

performance index

The performance indicators of block storage products mainly include IOPS、 Throughput and access latency. The performance of some block storage products is related to capacity, such as ESSD You need to select different capacity ranges for cloud disks.

I/O size
I/O Size refers to the amount of data for each read/write operation, such as 4 KiB。 Its performance indicators and those introduced later IOPS Operational relationship with throughput: IOPS * I/O Size=Throughput. So applied I/O The performance indicators that need to be concerned are different for different sizes.

IOPS (Input/Output Operations per Second): I/O The number indicates the read/write capacity of block storage, and the unit is times/second.

If you deploy the I/O It is characterized by random small delay I/O (such as database applications) needs attention IOPS Performance.

explain

In database applications, data insertion, update and deletion are very frequent. high IOPS It can ensure that the system can maintain efficient operation even under the pressure of a large number of random read and write operations, and avoid I/O Performance degradation or increased latency caused by bottlenecks.

frequently-used IOPS index

index	describe	Data access mode
total IOPS	Executed per second I/O Total number of operations	Discontinuous and continuous access to hard disk storage locations
random block read IOPS	Random reads performed per second I/O Average number of operations	Discontinuous access to hard disk storage locations
Random write IOPS	Random writes performed per second I/O Average number of operations	Discontinuous access to hard disk storage locations
Sequential reading IOPS	Sequential reads performed per second I/O Average number of operations	Continuous access to hard disk storage locations
Sequential writing IOPS	Sequential writes performed per second I/O Average number of operations	Continuous access to hard disk storage locations

Throughput: refers to the number of data that can be successfully transmitted per unit time, and the unit is MB/s。
If you deploy the I/O It is characterized by a large number of sequential reads and writes I/O (such as database applications) needs to focus on throughput.
explain
Hadoop Such offline computing business involves PB If the system throughput is low, the overall processing time will be very long, affecting business efficiency and response speed.
Access latency: block storage processes one I/O Time required, unit: s、ms Or μ s. Excessive delay will lead to application performance degradation or error reporting.
If your application is sensitive to high latency, such as database applications, you need to pay attention to latency. It is recommended that you use ESSD AutoPL Cloud disk ESSD Cloud disk and other low latency products.
Capacity: the size of storage space, in units of TiB、GiB、MiB perhaps KiB。
The block storage capacity is calculated in binary units, representing 1,024 The data size of carry, for example, 1 GiB=1024 MiB. Capacity is not a measure of the performance of block storage products, but different capacities can achieve different performance. The larger the capacity, the stronger the data processing capability of the storage device. Unit capacity of the same type of block storage products I/O The performance is consistent, but the cloud disk performance increases linearly with the capacity growth until it reaches the upper limit of the single disk performance of this type of block storage.

Cloud Disk Performance

The performance comparison of various cloud disks is shown in the following table.

important

The final performance of the cloud disk is limited not only by the size of the cloud disk itself, but also by the size of the attached instance. See storage I/O performance 。
SSD Cloud disk, efficient cloud disk and ordinary cloud disk belong to the previous generation of cloud disk products, and have gradually stopped selling in some regions and availability zones. Recommended ESSD PL0 Cloud disk or ESSD Entry Cloud disk replaces efficient cloud disk and ordinary cloud disk ESSD AutoPL Cloud Disk Replacement SSD Cloud disk.

Performance category	ESSD Series cloud disk								Previous generation cloud disk
Performance category	ESSD Local redundancy	ESSD AutoPL	ESSD PL-X (invited test)	ESSD Cloud disk PL3	ESSD Cloud disk PL2	ESSD Cloud disk PL1	ESSD Cloud disk PL0	ESSD Entry	SSD Cloud disk	Efficient cloud disk	Ordinary cloud disk
Single disk capacity range (GiB)	10~65,536	1~65,536	40~32,768	1,261~65,536	461~65,536	20~65,536	1~65,536	10~32,768	20~32,768	20~32,768	5~2,000
maximum IOPS	five ten thousand	one hundred ten thousand	three hundred ten thousand	one hundred ten thousand	ten ten thousand	five ten thousand	one ten thousand	6,000	25,000^②	5,000	hundreds
Maximum throughput (MB/s)	three hundred and fifty	4,096	12,288	4,000	seven hundred and fifty	three hundred and fifty	one hundred and eighty	one hundred and fifty	three hundred^②	one hundred and forty	30~40
Single disk IOPS Performance calculation formula^①	Min {1800+50 * capacity, 50000}	Benchmark performance: max {min {1800+50 * capacity, 50000}, 3000} Pre configured performance: Capacity (GiB)<=3: does not support setting preconfigured performance. Capacity (GiB)>=4: [1, min {(1000 IOPS/GiB * capacity base IOPS)，50,000}] Performance burst^③ : Actual Final IOPS Benchmark Performance IOPS preconfigured performance IOPS	Preconfigured IOPS	Min {1800+50 * capacity, 1000000}	Min {1800+50 * capacity, 100000}	Min {1800+50 * capacity, 50000}	Min {1800+12 * capacity, 10000}	Min {1800+8 * capacity, 6000}	Min {1800+30 * capacity, 25000}	Min {1800+8 * capacity, 5000}	nothing
Single disk throughput performance calculation formula (MB/s)^①	Min {120+0.5 * capacity, 350}	Benchmark performance: max {min {120+0.5 * capacity, 350}, 125} Pre configured performance 16 KB * pre configured IOPS/1,024 Performance burst^③ : Actual final throughput - Benchmark performance throughput - Preconfigured performance throughput	Min {4 KB * pre configured extra IOPS/1,024, 12,288}	Min {120+0.5 * capacity, 4000}	Min {120+0.5 * capacity, 750}	Min {120+0.5 * capacity, 350}	Min {100+0.25 * capacity, 180}	Min {100+0.15 * capacity, 150}	Min {120+0.5 * capacity, 300}	Min {100+0.15 * capacity, 140}	nothing
Data reliability	99.9999999%
Average delay of one-way random write (ms), Block Size=4K	millisecond ^④	zero point two	zero point zero three	zero point two	zero point two	zero point two	0.3~0.5	1~3	0.5~2	1~3	5~10

^① Description of single disk performance calculation formula:
- ESSD PL0 Cloud Disk Max IOPS Formulas: starting 1800 per GiB increase 12, up to 10,000。
- ESSD PL0 Maximum throughput formula of cloud disk: start 100 MB/s， each GiB increase 0.25 MB/s， Up to 180 MB/s。

^② SSD The performance of cloud disks varies according to the size of data blocks:

When IOPS When unchanged, the smaller the data block, the smaller the throughput.
When the throughput is constant, the smaller the data block, IOPS The larger.

I/O Size (KiB)	IOPS Maximum	Throughput (MB/s)
four	about 25,000	about one hundred
sixteen	about 17,200	about two hundred and sixty
thirty-two	about 9,600	about three hundred
sixty-four	about 4,800	about three hundred

^③ ESSD AutoPL In addition to the benchmark performance and preconfigured performance, cloud disks can also provide performance bursts. You can use the EBS CloudLens for EBS real-time monitoring ESSD AutoPL Cloud disk burst details, including Burst Time Burst Quantity (i.e. burst IO Total), etc. For more information, see Cloud Disk Analysis 。
^④ The data written to the ESSD local redundant cloud disk will be automatically distributed and stored in multiple zones, and RPO=0 will be achieved through physical replication. However, since data needs to be written to multiple different zones synchronously, the write latency varies in the availability zones of different regions, and is higher than ESSD PL1 Cloud disk. You can follow Test block storage performance , test the average write delay of local redundant cloud disks.

Local disk performance

warning

The local disk cannot be created separately. Its data reliability depends on the reliability of the physical machine. There is a single point of failure risk. A single point of failure of the physical machine may affect the operation of multiple instances. There is a risk of data loss when using the local disk to store data. Do not store business data that needs to be saved for a long time on the local disk. For a detailed introduction to the local disk, see Local disk 。

NVMe SSD Local disk

Big data instance type family d3c Carried NVMe SSD The local disk performance is shown in the following table.

index	Single disk performance	ecs.d3c.3xlarge	ecs.d3c.7xlarge	ecs.d3c.14xlarge
Maximum read IOPS	100,000	100,000	200,000	400,000
Maximum read throughput	4 GB/s	4 GB/s	8 GB/s	16 GB/s
Maximum write throughput	2 GB/s	2 GB/s	4 GB/s	8 GB/s
access delay	Microsecond (μ s)

local SSD Type Instance Type Family i4 Carried NVMe SSD The local disk performance is shown in the following table.

NVMe SSD index	ecs.i4.large	ecs.i4.xlarge	ecs.i4.2xlarge	ecs.i4.4xlarge	ecs.i4.8xlarge	ecs.i4.16xlarge	ecs.i4.32xlarge
Maximum read IOPS	eleven point two five ten thousand	twenty-two point five ten thousand	forty-five ten thousand	ninety ten thousand	one hundred and eighty ten thousand	three hundred and sixty ten thousand	seven hundred and twenty ten thousand
Maximum read throughput	0.75 GB/s	1.5 GB/s	3 GB/s	6 GB/s	12 GB/s	24 GB/s	48 GB/s
Maximum write throughput	0.375 GB/s	0.75 GB/s	1.5 GB/s	3 GB/s	6 GB/s	12 GB/s	24 GB/s
access delay	Microsecond (μ s)

explain

The indicators in the table are the best performance. It is recommended that you use the latest version of Linux Mirror (this instance type family only supports Linux Mirror) to maximize performance, such as Alibaba Cloud Linux 3 。

local SSD Type Instance Type Family i4g、i4r Carried NVMe SSD The local disk performance is shown in the following table.

NVMe SSD index	ecs.i4g.4xlarge and ecs.i4r.4xlarge	ecs.i4g.8xlarge and ecs.i4r.8xlarge	ecs.i4g.16xlarge and ecs.i4r.16xlarge	ecs.i4g.32xlarge and ecs.i4r.32xlarge
Maximum read IOPS	twenty-five ten thousand	fifty ten thousand	one hundred ten thousand	two hundred ten thousand
Maximum read throughput	1.5 GB/s	3 GB/s	6 GB/s	12 GB/s
Maximum write throughput	1 GB/s	2 GB/s	4 GB/s	8 GB/s
access delay	Microsecond (μ s)

explain

local SSD Type Instance Type Family i3 Carried NVMe SSD The local disk performance is shown in the following table.

NVMe SSD index	ecs.i3.xlarge	ecs.i3.2xlarge	ecs.i3.4xlarge	ecs.i3.8xlarge	ecs.i3.13xlarge	ecs.i3.26xlarge
Maximum read IOPS	twenty-five ten thousand	fifty ten thousand	one hundred ten thousand	two hundred ten thousand	three hundred ten thousand	six hundred ten thousand
Maximum read throughput	1.5 GB/s	3 GB/s	6 GB/s	12 GB/s	18 GB/s	36 GB/s
Maximum write throughput	1 GB/s	2 GB/s	4 GB/s	8 GB/s	12 GB/s	24 GB/s
access delay	Microsecond (μ s)

explain

local SSD Type Instance Type Family i3g Carried NVMe SSD The local disk performance is shown in the following table.

NVMe SSD index	ecs.i3g.2xlarge	ecs.i3g.4xlarge	ecs.i3g.8xlarge	ecs.i3g.13xlarge	ecs.i3g.26xlarge
Maximum read IOPS	twelve point five ten thousand	twenty-five ten thousand	fifty ten thousand	seventy-five ten thousand	one hundred and fifty ten thousand
Maximum read throughput	0.75 GB/s	1.5 GB/s	3 GB/s	4.5 GB/s	9 GB/s
Maximum write throughput	0.5 GB/s	1 GB/s	2 GB/s	3 GB/s	6 GB/s
access delay	Microsecond (μ s)

explain

local SSD Type Instance Type Family i2、i2g Carried NVMe SSD The local disk performance is shown in the following table.

NVMe SSD index	Single disk performance		Comprehensive instance performance^①
NVMe SSD index	only ecs.i2.xlarge and ecs.i2g.2xlarge	rest i2 and i2g Specifications	Comprehensive instance performance^①
Maximum capacity	894 GiB	1,788 GiB	8*1,788 GiB
Maximum read IOPS	fifteen ten thousand	thirty ten thousand	one hundred and fifty ten thousand
Maximum read throughput	1 GB/s	2 GB/s	16 GB/s
Maximum write throughput	0.5 GB/s	1 GB/s	8 GB/s
access delay	Microsecond (μ s)

^① The comprehensive instance performance is only applicable to ecs.i2.16xlarge， express i2 Local storage performance of the largest specification in the specification family.

local SSD Type Instance Type Family i2ne、i2gne Carried NVMe SSD The local disk performance is shown in the following table.

NVMe SSD index	ecs.i2ne.xlarge and ecs.i2gne.2xlarge	ecs.i2ne.2xlarge and ecs.i2gne.4xlarge	ecs.i2ne.4xlarge and ecs.i2gne.8xlarge	ecs.i2ne.8xlarge and ecs.i2gne.16xlarge	ecs.i2ne.16xlarge
Maximum capacity	894 GiB	1,788 GiB	2*1,788 GiB	4*1,788 GiB	8*1,788 GiB
Maximum read IOPS	twenty-five ten thousand	fifty ten thousand	one hundred ten thousand	two hundred ten thousand	four hundred ten thousand
Maximum read throughput	1.5 GB/s	3 GB/s	6 GB/s	12 GB/s	24 GB/s
Maximum write throughput	1 GB/s	2 GB/s	4 GB/s	8 GB/s	16 GB/s
access delay	Microsecond (μ s)

local SSD Type Instance Type Family i1 Carried NVMe SSD The local disk performance is shown in the following table.

NVMe SSD index	Single disk performance	Comprehensive instance performance^②
Maximum capacity	1,456 GiB	2,912 GiB
maximum IOPS	twenty-four ten thousand	forty-eight ten thousand
write IOPS^①	Min {165 * capacity, 240000}	2 * min {165 * capacity, 240000}
read IOPS^①	Min {165 * capacity, 240000}	2 * min {165 * capacity, 240000}
Maximum read throughput	2 GB/s	4 GB/s
Read Throughput^①	Min {1.4 * capacity, 2000} MB/s	2 * min {1.4 * capacity, 2000} MB/s
Maximum write throughput	1.2 GB/s	2.4 GB/s
Write Throughput^①	Min {0.85 * capacity, 1200} MB/s	2 * min {0.85 * capacity, 1200} MB/s
access delay	Microsecond (μ s)

^① Description of single disk performance calculation formula:

In single block NVMe SSD Local disk write IOPS Calculation formula description: every GiB by 165 IOPS， highest twenty-four ten thousand IOPS。
In single block NVMe SSD Description of the formula for calculating the write throughput of the local disk: every GiB by 0.85 MB/s， highest 1,200 MB/s。

^② The comprehensive instance performance is only applicable to ecs.i1.14xlarge， express i1 Local storage performance of the largest specification in the specification family.

SATA HDD Local disk

SATA HDD The local disk performance is shown in the following table.

SATA HDD index	d1、d1ne		d2c		d2s		d3s
SATA HDD index	Single disk performance	Comprehensive instance performance	Single disk performance	Comprehensive instance performance	Single disk performance	Comprehensive instance performance	Single disk performance	Comprehensive instance performance
Maximum capacity	5,500 GiB	154,000 GiB	3,700 GiB	44,400 GiB	7,300 GiB	219,000 GiB	11,100 GiB	355,200 GiB
Maximum throughput	190 MB/s	5,320 MB/s	190 MB/s	2,280 MB/s	190 MB/s	5,700 MB/s	260 MB/s	8,320 MB/s
access delay	Milliseconds

explain

The comprehensive instance performance is only applicable to ecs.d1.14xlarge、ecs.d1ne.14xlarge、ecs.d2c.24xlarge、ecs.d2s.20xlarge、ecs.d3s.16xlarge Specification, which represents the local storage performance of the maximum specification in each specification family.

Performance of elastic temporary disk

explain

The flexible temporary disk can be used as temporary data storage. For an introduction to flexible temporary disks, see Elastic temporary disk 。

The standard version is more suitable for scenarios with large data volume and high throughput requirements, while the advanced version is more suitable for scenarios with small capacity but high throughput requirements IOPS High demand scenarios. The performance that can be achieved by different types is described in the following table:

Performance category	Standard elastic temporary disk	Advanced elastic temporary disk
Single disk capacity range (GiB)	64~8,192	64~8,192
Single disk maximum reading IOPS	min^② {100 * capacity, 820000}	Min {300 * capacity, 1000000}
Single disk maximum write IOPS	Min {20 * capacity, 160000}	Min {150 * capacity, 500000}
Maximum read throughput of a single disk (MB/s)	Min {0.8 * capacity, 4096}	Min {1.6 * capacity, 4096}
Maximum write throughput of a single disk (MB/s)	Min {0.4 * capacity, 2048}	Min {1 * capacity, 2048}
write I/O density^①	twenty	one hundred and fifty
read I/O density^①	one hundred	three hundred

^① ：IO Density=IOPS/cloud disk capacity, unit: IOPS/GiB， Represents every GiB Reachable IOPS Ability.

^② ：min{A，B}， Indicates return A and B The smaller of the two values.

test Block storage performance

You can refer to the following methods to test the performance of block storage:

Cloud disk read/write slow/IO High troubleshooting ideas

You can use the ECS Console EBS The console or cloud monitoring console views various monitoring information of the cloud disk, and judges whether the current cloud disk performance meets business needs or reaches performance bottlenecks. For more information, see View cloud disk monitoring information 。

Confirm whether the cloud disk billing method is volume based. If pay as you go, when the account is in arrears, the disk IO The speed will be limited, and the account can be restored after recharging.
be careful: Arrearage fifteen The cloud disk that has not been recharged and renewed in days will be automatically released, and the data in it cannot be retrieved , please pay attention.
Linux The system can refer to Linux System Disk I/O Troubleshooting and solutions for high load problems , Troubleshooting IOPS High occupancy programs.
When importing data, the performance of both the client and server will affect the read/write speed.
Can be used in the server use atop Tool monitoring Linux System indicators , this tool can continuously monitor the use of various resources in the server, and the resource use information will be recorded in/var/log/top by default In the directory, you can use atop Logs help to further locate problem points.
If the cloud disk performance does not meet the business requirements, you can also try to improve the cloud disk performance. Please refer to How to improve cloud disk performance 。

How to improve cloud disk performance

If the current cloud disk performance does not meet the business requirements, you can try the following methods to improve the cloud disk performance:

important

The final performance of the cloud disk is not only limited by the size of the cloud disk itself, but also by the size of the attached instance IOPS When the bandwidth is lower than the performance limit of the cloud disk itself, improving the cloud disk performance alone cannot change the final performance limit of the cloud disk. You need to upgrade the instance type to improve it. For the limitation of the instance type on the cloud disk, you can refer to Instance Type Family 。

Application scenarios	Methods to improve performance
When the cloud disk type (such as SSD Cloud disk) cannot meet the higher requirements brought by business growth IOPS Or throughput requirements, you can choose to change to a higher performance cloud disk type, such as ESSD PL1， To achieve higher IOPS And better response time. This method is suitable for applications that have strict requirements on storage performance and have significant growth in business scale or access.	Change cloud disk type
If you are using ESSD Cloud disk, you can adjust according to the change of business load ESSD The performance level of the cloud disk.	modify ESSD Cloud Disk Performance Level
If you are using ESSD AutoPL Cloud disk, you can set preconfigured performance or enable performance burst to improve the performance of cloud disk.	modify ESSD AutoPL Cloud Disk Performance Configuration
When your business needs more than IOPS， At the same time, the storage space is also insufficient. It is recommended to expand the cloud disk. For some types of cloud disks (such as ESSD PL1）， As capacity increases, the foundation IOPS It will also be improved accordingly, so as to enhance the processing capacity and performance of cloud disks. It is suitable for the continuous growth of data volume, and for the storage capacity and IOPS All have high requirements for applications. for example ESSD PL1 Cloud disk IOPS=min {1800+50 * capacity, 50000}, 40 GiB Of ESSD PL1 Cloud disk IOPS by 3800, expanded to 100 GiB When, IOPS by 6,800。	Expand cloud disk
When you need to manage and optimize storage resource allocation more flexibly and want to improve cloud disk performance, you can choose LVM。 By distributing data on multiple logical volumes, you can achieve parallel processing of read and write operations, thereby improving the performance of cloud disk, especially for scenarios requiring high concurrent access such as multithreaded applications and databases.	Create logical volume
When you need to improve data redundancy IOPS And throughput, you can choose to create RAID Array. For example, using RAID 0 It can improve reading and writing speed, while RAID 1 or RAID 10 It can provide data redundancy while improving performance.	establish RAID array