Edit

Azure Database for MySQL - Flexible Server service tiers

  • Article

APPLIES TO: Azure Database for MySQL - Flexible Server

You can create an Azure Database for MySQL flexible server instance in one of three different service tiers: Burstable, General Purpose, and Business Critical. The service tiers are differentiated by the underlying VM SKU used B-series, D-series, and E-series. The choice of compute tier and size determines the memory and vCores available on the server. The same storage technology is used across all service tiers. All resources are provisioned at the Azure Database for MySQL flexible server instance level. A server can have one or many databases.

Resource / Tier Burstable General Purpose Business Critical
VM series B-series Dadsv5-seriesDdsv4-series Edsv4/Edsv5-series*/Eadsv5-series
vCores 1, 2, 4, 8, 12, 16, 20 2, 4, 8, 16, 32, 48, 64 2, 4, 8, 16, 32, 48, 64, 80, 96
Memory per vCore Variable 4 GiB 8 GiB **
Storage size 20 GiB to 16 TiB 20 GiB to 16 TiB 20 GiB to 16 TiB
Database backup retention period 1 to 35 days 1 to 35 days 1 to 35 days

** With the exception of 64,80, and 96 vCores, which has 504 GiB, 504 GiB and 672 GiB of memory respectively.

* Ev5 compute provides best performance among other VM series in terms of QPS and latency. Learn more about performance and region availability of Ev5 compute from here.

To choose a compute tier, use the following table as a starting point.

Compute tier Target workloads
Burstable Best for workloads that don’t need the full CPU continuously.
General Purpose Most business workloads that require balanced compute and memory with scalable I/O throughput. Examples include servers for hosting web and mobile apps and other enterprise applications.
Business Critical High-performance database workloads that require in-memory performance for faster transaction processing and higher concurrency. Examples include servers for processing real-time data and high-performance transactional or analytical apps.

After you create a server, the compute tier, compute size, and storage size can be changed. Compute scaling requires a restart and takes between 60-120 seconds, while storage scaling doesn't require restart. You also can independently adjust the backup retention period up or down. For more information, see the Scale resources section.

Service tiers, size, and server types

Compute resources can be selected based on the tier and size. This determines the vCores and memory size. vCores represent the logical CPU of the underlying hardware. The detailed specifications of the available server types are as follows:

Compute size vCores Physical Memory Size (GiB) Total Memory Size (GiB) Max Supported IOPS Max Connections Temp Storage (SSD) GiB
Burstable
Standard_B1s 1 1 1.1 320 171 0
Standard_B1ms 1 2 2.2 640 341 0
Standard_B2s 2 4 4.4 1280 683 0
Standard_B2ms 2 8 8.8 1700 1365 0
Standard_B4ms 4 16 17.6 2400 2731 0
Standard_B8ms 8 32 35.2 3100 5461 0
Standard_B12ms 12 48 52.8 3800 8193 0
Standard_B16ms 16 64 70.4 4300 10923 0
Standard_B20ms 20 80 88 5000 13653 0
General Purpose
Standard_D2ads_v5 2 8 11 3200 1365 53
Standard_D2ds_v4 2 8 11 3200 1365 53
Standard_D4ads_v5 4 16 22 6400 2731 107
Standard_D4ds_v4 4 16 22 6400 2731 107
Standard_D8ads_v5 8 32 44 12800 5461 215
Standard_D8ds_v4 8 32 44 12800 5461 215
Standard_D16ads_v5 16 64 88 20000 10923 430
Standard_D16ds_v4 16 64 88 20000 10923 430
Standard_D32ads_v5 32 128 176 20000 21845 860
Standard_D32ds_v4 32 128 176 20000 21845 860
Standard_D48ads_v5 48 192 264 20000 32768 1290
Standard_D48ds_v4 48 192 264 20000 32768 1290
Standard_D64ads_v5 64 256 352 20000 43691 1720
Standard_D64ds_v4 64 256 352 20000 43691 1720
Business Critical
Standard_E2ds_v4 2 16 22 5000 2731 37
Standard_E2ads_v5 2 16 22 5000 2731 37
Standard_E4ds_v4 4 32 44 10000 5461 75
Standard_E4ads_v5 4 32 44 10000 5461 75
Standard_E8ds_v4 8 64 88 18000 10923 151
Standard_E8ads_v5 8 64 88 18000 10923 151
Standard_E16ds_v4 16 128 176 28000 21845 302
Standard_E16ads_v5 16 128 176 28000 21845 302
Standard_E20ds_v4 20 160 220 28000 27306 377
Standard_E20ads_v5 20 160 220 28000 27306 377
Standard_E32ds_v4 32 256 352 38000 43691 604
Standard_E32ads_v5 32 256 352 38000 43691 604
Standard_E48ds_v4 48 384 528 48000 65536 906
Standard_E48ads_v5 48 384 528 48000 65536 906
Standard_E64ds_v4 64 504 693 64000 86016 1224
Standard_E64ads_v5 64 504 693 64000 86016 1224
Standard_E80ids_v4 80 504 693 72000 86016 1224
Standard_E2ds_v5 2 16 22 5000 2731 37
Standard_E4ds_v5 4 32 44 10000 5461 75
Standard_E8ds_v5 8 64 88 18000 10923 151
Standard_E16ds_v5 16 128 176 28000 21845 302
Standard_E20ds_v5 20 160 220 28000 27306 377
Standard_E32ds_v5 32 256 352 38000 43691 604
Standard_E48ds_v5 48 384 528 48000 65536 906
Standard_E64ds_v5 64 512 704 64000 87383 1208
Standard_E96ds_v5 96 672 924 80000 100000 2004

Memory management in Azure Database for MySQL Flexible Server

In MySQL, memory plays an important role throughout various operations, including query processing and caching. Azure Database for MySQL flexible server optimizes memory allocation for the MySQL server process (mysqld), ensuring it receives sufficient memory resources for efficient query processing, caching, client connection management, and thread handling. Learn more on how MySQL uses memory.

Physical Memory Size (GB)

The Physical memory Size (GB) in the table below represents the available random-access memory (RAM) in gigabytes (GB) on your Azure Database for MySQL flexible server.

Total Memory Size (GB)

Azure Database for MySQL flexible server provides a Total Memory Size (GB). This represents the total memory available to your server, which is a combination of physical memory and a set amount of temporary storage SSD component. This unified view is designed to streamline resource management, allowing you to focus on the total memory available to your Azure MySQL Server (mysqld) process only. Memory Percent (memory_percent) metric represents the percentage of memory occupied by the Azure MySQL server process (mysqld). This metric is calculated from the Total Memory Size (GB). For example, when the Memory Percent metric displays a value of 60, it means that your Azure MySQL Server process is utilizing 60% of the Total memory size (GB) available on your Azure Database for MySQL flexible server.

MySQL Server (mysqld)

The Azure MySQL server process, mysqld, serves as the core engine for database operations. Upon startup, it initializes total components such as the InnoDB buffer pool and thread cache, utilizing memory based on configuration and workload demands. For example, the InnoDB buffer pool caches frequently accessed data and indexes to improve query execution speed, while the thread cache manages client connection threads. Learn more.

InnoDB Storage Engine

As MySQL's default storage engine, InnoDB uses memory for caching frequently accessed data and managing internal structures like the innodb buffer pool and log buffer. InnoDB buffer pool holds table data and indexes in memory to minimize disk I/O, enhancing performance. InnoDB Buffer Pool Size parameter is calculated based on the physical memory size (GB) available on server. Learn more on the sizes of the InnoDB Buffer Pool available in Azure Database for MySQL flexible server.

Threads

Client connections are managed through dedicated threads handled by the connection manager. These threads handle authentication, query execution, and result retrieval for client interactions. Learn more.

To get more details about the compute series available, refer to Azure VM documentation for Burstable (B-series), General Purpose Dadsv5-seriesDdsv4-series, and Business Critical Edsv4/Edsv5-series/Eadsv5-series.

Performance limitations of burstable series instances

Note

For Burstable (B-series) compute tier if the VM is started/stopped or restarted, the credits may be lost. For more information, see Burstable (B-Series) FAQ.

Burstable compute tier is designed to provide a cost-effective solution for workloads that don't require continuous full CPU continuously. This tier is ideal for nonproduction workloads, such as development, staging, or testing environments. The unique feature of the burstable compute tier is its ability to “burst”, that is, to utilize more than its baseline CPU performance using up to 100% of the vCPU when the workload requires it. This is made possible by a CPU credit model, which allows B-series instances to accumulate “CPU credits” during periods of low CPU usage. These credits can then be spent during periods of high CPU usage, allowing the instance to burst above its base CPU performance.

However, it’s important to note that once a burstable instance exhausts its CPU credits, it operates at its base CPU performance. For example, the base CPU performance of a Standard_B1ms is 20% that is, 0.2 Vcore. If Burstable tier server is running a workload that requires more CPU performance than the base level, and it has exhausted its CPU credits, the server may experience performance limitations and eventually could affect various system operations like Stop/Start/Restart for your server.

Note

For servers in Burstable (B-series) compute tier, such as Standard_B1s/Standard_B1ms/Standard_B2s, their relatively smaller host memory size, may lead to crashes (out of memory) under continouous workload, even if the memory_percent metric has not reached 100%.

Due to this throttling, the server may encounter connectivity issues and system operations might be affected. In such situations, the recommended course of action is to pause the workload on the server to accumulate credits according to the B-series credit banking model, or to consider scaling the server to higher tiers such as General Purpose or Business Critical tiers.

Therefore, while the Burstable compute tier offers significant cost and flexibility advantages for certain types of workloads, it is not recommended for production workloads that require consistent CPU performance. The Burstable tier doesn't support functionality of creating Read Replicas and High availability feature. For such workloads and features, other compute tiers, such as the General Purpose or Business Critical are more appropriate.

For more information on the Azure's B-series CPU credit model, see the B-series burstable instances and B-series CPU credit model.

Monitoring CPU credits in burstable tier

Monitoring your CPU credit balance is crucial for maintaining optimal performance in the Burstable compute tier. Azure Database for MySQL Flexible Server provides two key metrics related to CPU credits. The ideal threshold for triggering an alert depends on your specific workload and performance requirements.

CPU Credit Consumed: This metric indicates the number of CPU credits consumed by your instance. Monitoring this metric can help you understand your instance’s CPU usage patterns and manage its performance effectively.

CPU Credit Remaining: This metric shows the number of CPU credits remaining for your instance. Keeping an eye on this metric can help you prevent your instance from degrading in performance due to exhausting its CPU credits. If the CPU Credit Remaining metric drops below a certain level (for example, less than 30% of the total available credits), this would indicate that the instance is at risk of exhausting its CPU credits if the current CPU load continues.

For more information, on how to setup alerts on metrics, refer to this guide.

Storage

The storage you provision is the amount of storage capacity available to your flexible server. Storage is used for the database files, temporary files, transaction logs, and the MySQL server logs. In all service tiers, the minimum storage supported is 20 GiB and maximum is 16 TiB. Storage is scaled in 1-GiB increments and can be scaled up after the server is created.

Note

Storage can only be scaled up, not down.

You can monitor your storage consumption in the Azure portal (with Azure Monitor) using the storage limit, storage percentage, and storage used metrics. Refer to the monitoring article to learn about metrics.

Reaching the storage limit

When storage consumed on the server is close to reaching the provisioned limit, the server is put to read-only mode to protect any lost writes on the server. Servers with less than equal to 100 GiB provisioned storage are marked read-only if the free storage is less than 5% of the provisioned storage size. Servers with more than 100 GiB provisioned storage are marked read only when the free storage is less than 5 GiB.

For example, if you have provisioned 110 GiB of storage, and the actual utilization goes over 105 GiB, the server is marked read-only. Alternatively, if you have provisioned 5 GiB of storage, the server is marked read-only when the free storage reaches less than 256 MB.

While the service attempts to make the server read-only, all new write transaction requests are blocked and existing active transactions continue to execute. When the server is set to read-only, all subsequent write operations and transaction commits fail. Read queries continue to work uninterrupted.

To get the server out of read-only mode, you should increase the provisioned storage on the server. This can be done using the Azure portal or Azure CLI. Once increased, the server is ready to accept write transactions again.

We recommended that you set up an alert to notify you when your server storage is approaching the threshold so you can avoid getting into the read-only state. For more information, see the documentation on alert documentation how to set up an alert.

Storage auto grow

Storage autogrow prevents your server from running out of storage and becoming read-only. If storage autogrow is enabled, the storage automatically grows without impacting the workload. Storage autogrow is enabled by default for all new server creates. For servers with less than equal to 100 GB provisioned storage, the provisioned storage size is increased by 5 GB when the free storage is below 10% of the provisioned storage. For servers with more than 100 GB of provisioned storage, the provisioned storage size is increased by 5% when the free storage space is below 10 GB of the provisioned storage size. Maximum storage limits as specified above apply. Refresh the server instance to see the updated storage provisioned under Settings on the Compute + Storage page.

For example, if you have provisioned 1,000 GB of storage, and the actual utilization goes over 990 GB, the server storage size is increased to 1,050 GB. Alternatively, if you have provisioned 20 GB of storage, the storage size is increase to 25 GB when less than 2 GB of storage is free.

Remember that storage once autoscaled up, can't be scaled down.

Note

Storage autogrow is default enabled for a High-Availability configured server and can not to be disabled.

IOPS

Azure Database for MySQL flexible server supports pre-provisioned IOPS and autoscale IOPS. Learn more. The minimum IOPS are 360 across all compute sizes and the maximum IOPS is determined by the selected compute size. To learn more about the maximum IOPS per compute size refer to the table.

Important

**Minimum IOPS are 360 across all compute sizes
**Maximum IOPS are determined by the selected compute size.

You can monitor your I/O consumption in the Azure portal (with Azure Monitor) using IO percent metric. If you need more IOPS than the max IOPS based on compute, then you need to scale your server's compute.

Pre-provisioned IOPS

Azure Database for MySQL flexible server offers pre-provisioned IOPS, allowing you to allocate a specific number of IOPS to your Azure Database for MySQL flexible server instance. This setting ensures consistent and predictable performance for your workloads. With pre-provisioned IOPS, you can define a specific IOPS limit for your storage volume, guaranteeing the ability to handle some requests per second. This results in a reliable and assured level of performance. Pre-provisioned IOPS enables you to provision additional IOPS above the IOPS limit. Using this feature, you can increase or decrease the number of IOPS provisioned based on your workload requirements at any time.

Autoscale IOPS

The cornerstone of Azure Database for MySQL flexible server is its ability to achieve the best performance for tier 1 workloads, which can be improved by enabling server automatically scale performance (IO) of its database servers seamlessly depending on the workload needs. This is an opt-in feature that enables users to scale IOPS on demand without having to pre-provision a certain amount of IO per second. With the Autoscale IOPS featured enable, you can now enjoy worry free IO management in Azure Database for MySQL flexible server because the server scales IOPs up or down automatically depending on workload needs.

With Autoscale IOPS, you pay only for the IO the server use and no longer need to provision and pay for resources they aren’t fully using, saving both time and money. In addition, mission-critical Tier-1 applications can achieve consistent performance by making additional IO available to the workload at any time. Autoscale IOPS eliminate the administration required to provide the best performance at the least cost for Azure Database for MySQL flexible server customers.

Dynamic Scaling: Autoscale IOPS dynamically adjust the IOPS limit of your database server based on the actual demand of your workload. This ensures optimal performance without manual intervention or configuration.

Handling Workload Spikes: Autoscale IOPS enable your database to seamlessly handle workload spikes or fluctuations without compromising the performance of your applications. This feature ensures consistent responsiveness even during peak usage periods.

Cost Savings: Unlike the Pre-provisioned IOPS where a fixed IOPS limit is specified and paid for regardless of usage, Autoscale IOPS lets you pay only for the number of I/O operations that you consume.

Backup

The service automatically takes backups of your server. You can select a retention period from a range of 1 to 35 days. Learn more about backups in the backup and restore concepts article.

Scale resources

After you create your server, you can independently change the compute tier, compute size (vCores and memory), and the amount of storage, and the backup retention period. The compute size can be scaled up or down. The backup retention period can be scaled up or down from 1 to 35 days. The storage size can only be increased. Scaling of the resources can be done through the portal or Azure CLI.

Note

The storage size can only be increased. You cannot go back to a smaller storage size after the increase.

When you change the compute tier or compute size, the server is restarted for the new server type to take effect. During the moment when the system switches over to the new server, no new connections can be established, and all uncommitted transactions are rolled back. This window varies, but in most cases, is between 60-120 seconds.

Scaling storage and changing the backup retention period are online operations and don't require a server restart.

Pricing

For the most up-to-date pricing information, see the service pricing page. To see the cost for the configuration you want, the Azure portal shows the monthly cost on the Compute + storage tab based on the options you select. If you don't have an Azure subscription, you can use the Azure pricing calculator to get an estimated price. On the Azure pricing calculator website, select Add items, expand the Databases category, choose Azure Database for MySQL, and Flexible Server as the deployment type to customize the options.

If you would like to optimize server cost, you can consider following tips:

  • Scale down your compute tier or compute size (vCores) if compute is underutilized.
  • Consider switching to the Burstable compute tier if your workload doesn't need the full compute capacity continuously from the General Purpose and Business Critical tiers.
  • Stop the server when not in use.
  • Reduce the backup retention period if a longer retention of backup isn't required.

Next steps