All Products
Search
Document Center

Elastic Compute Service:Overview

Last Updated:Aug 21, 2024

Burstable instances are an economical instance type that is designed to meet performance burst requirements in entry-level computing scenarios. This topic describes the features, usage scenarios, instance families, instance types, baseline performance, CPU credits, and performance modes of burstable instances.

Introduction

Burstable instances use CPU credits to maintain computing performance and are suitable for scenarios in which CPU utilization is usually low but suddenly increases. Burstable instances can accumulate CPU credits and consume the credits to burst performance above the baseline based on your workload requirements. This consumption pattern does not affect the environments or applications that are deployed on the burstable instances. Burstable instances are more flexible and cost-effective than other types of instances in terms of CPU utilization.

The CPU credit mechanism allows you to save computing power during off-peak hours for use during peak hours and reduce costs. If you have unplanned performance requirements, you can enable the unlimited mode for the burstable instances.

The following burstable instance families are available:

Note

Burstable instances are special shared instances. For information about shared instance families, see Shared instance families.

The following table describes the baseline performance, CPU credits, and performance modes of burstable instances.

Term

Description

References

baseline performance

The amount of CPU capacity that is continuously provisioned to a burstable instance. The baseline performance varies based on the instance type.

Baseline performance

initial CPU credit

A limited number of CPU credits that are allocated to a new burstable instance. Thirty initial CPU credits are allocated to each vCPU.

CPU credits

CPU credit balance

The total number of net CPU credits that are accrued when the earned CPU credits exceed the consumed CPU credits. You can use the CPU credit balance to run burstable instances above the performance baseline.

CPU credits

max CPU credit balance

The maximum number of CPU credits that can be earned by a burstable instance within a 24-hour period. The CPU credit balance is valid for 24 hours. Each burstable instance earns CPU credits at a specific rate based on the instance type and can accrue only a limited number of credits in the CPU credit balance.

CPU credits

performance mode

A burstable instance can run in standard mode or unlimited mode.

  • A burstable instance in standard mode cannot burst above the performance baseline if the available CPU credits are depleted.

  • A burstable instance in unlimited mode can sustain CPU utilization above the baseline for a period of time by consuming overdrawn or advance CPU credits. You may be charged for the CPU credits.

Performance modes

advance CPU credit

The CPU credits that a burstable instance earns over the next 24 hours. You may be charged for advance CPU credits. Advance CPU credits can be consumed only if the unlimited mode is enabled.

Performance modes

overdrawn CPU credit

The CPU credits that a burstable instance consumes after the instance depletes the advance credits to remain running above the performance baseline. You are charged for overdrawn CPU credits. Overdrawn CPU credits can be consumed only if the unlimited mode is enabled.

Performance modes

Scenarios

When you purchase an enterprise-level instance, the vCPUs of the instance are exclusively reserved for your use. You are charged for the vCPUs regardless of whether the vCPUs run at 100% utilization. If you require a high level of CPU performance for only a specific time of the day, specific vCPU resources are left idle for the remainder of the day, but you are still charged for the unused resources. To prevent this issue, you can use burstable instances to better meet your business requirements.

Burstable instances are suitable for scenarios that require higher-than-normal performance for a specific period of time, such as stress testing service applications, lightweight applications, microservices, and web application servers. We recommend that you evaluate your business requirements to determine the performance levels required during off-peak and peak hours before you make a purchase. Purchase burstable instances whose baseline performance meets your business requirements during off-peak hours. This way, you can achieve the required performance at significantly lower costs.

Note
  • If the purchased burstable instances do not meet your business requirements, you can change the instance type. For more information, see the Instance type changes section of this topic.

  • Windows application workloads and GUIs require a high level of CPU performance. If you run Windows application workloads or GUIs on an instance of a t-series burstable instance type, the instance may run slowly or go down. We recommend that you change the instance type based on your business requirements. For example, you can change the instance type to a g-series general-purpose instance type, c-series compute-optimized instance type, or r-series memory-optimized instance type.

Baseline performance

Baseline performance is the amount of CPU capacity that is continuously provisioned to a burstable instance. The baseline performance varies based on the instance type. You can view the baseline performance of different instance types in the Average baseline CPU performance column of the instance type tables.

CPU credits

CPU credits are the computing resources that are available for your use. The computing resources determine the computing performance that burstable instances can deliver. The following section describes the terms and examples related to CPU credits:

  • initial CPU credit

    When you create a burstable instance, 30 CPU credits are allocated to each vCPU of the instance, which are the initial CPU credits. The initial CPU credits allow you to complete deployment tasks when you start the instance.

    For example, an ecs.t5-lc1m2.large instance has two vCPUs and is allocated 60 initial CPU credits when the instance is created. An ecs.t5-c1m1.xlarge instance has four vCPUs and is allocated 120 initial CPU credits when the instance is created.

  • rate of earning CPU credits

    When you start a burstable instance, the instance begins to consume CPU credits to maintain the computing performance of the instance. At the same time, the burstable instance earns CPU credits at a specific rate that is determined by the instance type. The number of CPU credits that a vCPU can earn per hour varies based on the instance type. The CPU credits per hour column in the instance type tables indicates the CPU credits that all vCPUs of a burstable instance can earn per hour.

    For example, a baseline performance of 25% of an ecs.t5-c1m1.large instance indicates that the CPU credits that a vCPU of the instance earns per hour can keep the vCPU running at 25% utilization for 1 hour or at 100% utilization for 15 minutes (60 × 25%). Based on the baseline performance of the burstable instance, each vCPU earns 15 CPU credits per hour. An ecs.t5-c1m1.large instance has two vCPUs and earns 30 CPU credits per hour.

  • CPU credit balance

    If the earned CPU credits exceed the consumed CPU credits, the net credits are accrued as CPU credit balance. The CPU credit balance is valid for 24 hours. Each burstable instance earns CPU credits at a specific rate based on the instance type and can accrue only a limited number of credits in the CPU credit balance. The maximum CPU credit balance for a specific instance type is the maximum number of CPU credits that the burstable instance can earn within a 24-hour period. For more information, see the Max CPU credit balance column in the instance type tables.

    For example, an ecs.t5-c1m1.large instance can earn 30 CPU credits per hour. The maximum CPU credit balance that the instance can earn is 720 (30 × 24) credits.

  • rate of consuming CPU credits

    The rate at which a burstable instance consumes CPU credits varies based on the number of vCPUs, CPU utilization, and operating hours of the instance. For example, one CPU credit is consumed in the following scenarios:

    • One vCPU runs at 100% utilization for 1 minute.

    • One vCPU runs at 50% utilization for 2 minutes.

    • Two vCPUs run at 25% utilization for 2 minutes.

    When you start a burstable instance, the instance begins to consume CPU credits to maintain the computing performance of the instance. Initial credits that cannot be replenished are consumed first. When the initial credits are depleted, the burstable instance continues to consume the accrued CPU credits.

    • When the CPU utilization is below the baseline, the CPU credits earned are more than the CPU credits consumed and the CPU credit balance increases.

    • When the CPU utilization is equal to the baseline, the CPU credits earned are equal to the CPU credits consumed and the CPU credit balance remains unchanged.

    • When the CPU utilization is higher than the baseline, the CPU credits earned are fewer than the CPU credits consumed and the CPU credit balance decreases.

Note

The CPU utilization data collected from physical machines include the simulation overhead of privileged instructions. CPU utilization data is used to calculate the CPU credit usage. To view the CPU utilization of an instance, log on to the CloudMonitor console, click Host Monitoring in the left-side navigation pane, and then click the ID of an Elastic Compute Service (ECS) instance on the Host Monitoring page. Then, click the Basic Monitoring tab to view the CPU utilization of the instance. For more information, see Overview.

In different scenarios, stopping instances may have different impacts on your CPU credits:

  • If a pay-as-you-go instance is stopped in standard mode, the CPU credit balance of the instance is retained and the instance continues to earn CPU credits.

  • If a pay-as-you-go instance is stopped in economical mode, the CPU credit balance of the instance becomes invalid and the instance cannot continue to earn credits. When the pay-as-you-go instance is restarted, the instance receives initial CPU credits and begins to earn CPU credits.

  • If a pay-as-you-go instance is stopped due to an overdue payment, the CPU credit balance of the instance is retained but the instance cannot continue to earn credits until you complete the payment.

  • If a subscription instance expires and is stopped, the CPU credit balance of the instance is retained but the instance cannot continue to earn CPU credits. When the subscription instance is restarted, the instance begins to earn CPU credits.

Performance modes

A burstable instance can run in standard mode or unlimited mode.

  • Standard mode

    The performance of a burstable instance in standard mode is limited by the availability of CPU credits. After a burstable instance depletes all initial CPU credits and CPU credit balance, the instance cannot run above the performance baseline. When the CPU credit balance is low, the instance gradually reduces performance to the performance baseline within 15 minutes. This way, the instance does not experience a sudden decrease in performance when its CPU credit balance is depleted.

    The standard mode is suitable for scenarios such as lightweight web servers, development and testing environments, and databases that have low and medium performance. In the preceding scenarios, workloads are stable, instances do not run above the baseline performance for an extended period of time, and performance bursts are only occasionally required.

  • Unlimited mode

    The performance of a burstable instance in unlimited mode is not limited by the availability of CPU credits. You can overdraw or pay for additional CPU credits to meet performance burst requirements. If a burstable instance continues to run above the performance baseline after the initial CPU credits and the accrued credits are depleted, the instance begins to consume advance CPU credits and overdrawn CPU credits.

    • Advance CPU credits: Advance CPU credits are the CPU credits that a burstable instance earns over the next 24-hour period. You may be charged for advance CPU credits.

    • Overdrawn CPU credits: Overdrawn CPU credits are the CPU credits that a burstable instance consumes after the instance depletes the advance credits to remain running above the performance baseline. You are charged for overdrawn credits.

    Note

    For information about the fees and billing rules for advance CPU credits and overdrawn CPU credits, see the Additional fees section of the "Billing" topic.

    The following figure shows how the CPU credits change when a burstable instance runs in unlimited mode.

    image
    Note

    If the burstable instance consumed advance CPU credits and is stopped in economical mode, is released, experienced configurations changes, or switches to standard mode before the advance CPU credits are replenished, you are charged a lump sum amount for the advance CPU credits that are consumed.

    If you want to consume advance or overdrawn CPU credits in addition to your credit balance to meet performance burst requirements, you can enable unlimited mode for your burstable instances. Examples:

    • Events such as new feature releases, e-commerce promotions, and website promotions cause a substantial increase in your workloads. High CPU performance is required during the events. In this case, you can enable the unlimited mode for your burstable instances. You can disable the unlimited mode to reduce costs when the workload returns to normal.

    • Specific web applications may require performance bursts during a specific period of time during the day, but the daily average CPU utilization is below the baseline. In this case, you can enable the unlimited mode for your burstable instances during peak hours to ensure positive user experience. If the CPU credits that are earned during off-peak hours can offset the advance CPU credits that are consumed during peak hours, you can continue to provide positive user experience at no additional costs.

By default, the standard mode is enabled when you create a burstable instance. For information about how to enable the unlimited mode, see the Enable the unlimited mode section of the "Enable or disable the unlimited mode" topic.

For information about how CPU credits change when you run a burstable instance in different performance modes, see CPU credit change examples.

Instance type changes

When you monitor a burstable instance, you may find that the CPU utilization remains above or below the baseline for an extended period of time. This indicates that the instance type cannot meet your business requirements. We recommend that you re-evaluate the instance type to determine whether to change the instance type. The operation that you can perform to change instance types varies based on the instance billing method. For more information, see Overview of instance configuration changes.

t6, burstable instance family

Features:

  • Provides a CPU performance baseline and the ability to burst above the baseline, which are governed by accrued CPU credits.

  • More cost-effective compared with the t5 burstable instance family.

  • Compute:

    • Uses 2.5 GHz Intel® Xeon® Cascade Lake processors that deliver a turbo frequency of 3.2 GHz.

    • Uses DDR4 memory.

  • Storage:

    • Supports Enterprise SSDs (ESSDs), ESSD AutoPL disks, standard SSDs, and ultra disks.

      Note

      ESSDs at performance level (PL) 2 and 3 cannot provide maximum performance due to the specification limits of burstable instances. We recommend that you use enterprise-level instances or ESSDs of lower performance levels.

  • Network:

    • Supports IPv6.

    • Supports only virtual private clouds (VPCs).

  • Supported scenarios:

    • Web application servers

    • Lightweight applications and microservices

    • Development and testing environments

Instance types

Instance type

vCPU

Memory (GiB)

Average baseline CPU performance

CPU credits per hour

Max CPU credit balance

Network baseline/burst bandwidth (Gbit/s)

Packet forwarding rate (pps)

NIC queues

ENIs

Private IPv4 addresses per ENI

ecs.t6-c4m1.large

2

0.5

5%

6

144

0.08/burstable up to 0.4

40,000

1

2

2

ecs.t6-c2m1.large

2

1.0

10%

12

288

0.08/burstable up to 0.6

60,000

1

2

2

ecs.t6-c1m1.large

2

2.0

20%

24

576

0.08/burstable up to 1

100,000

1

2

2

ecs.t6-c1m2.large

2

4.0

20%

24

576

0.08/burstable up to 1

100,000

1

2

2

ecs.t6-c1m4.large

2

8.0

30%

36

864

0.08/burstable up to 1

100,000

1

2

2

ecs.t6-c1m4.xlarge

4

16.0

40%

96

2304

0.16/burstable up to 2

200,000

1

2

6

ecs.t6-c1m4.2xlarge

8

32.0

40%

192

4608

0.32/burstable up to 4

400,000

1

2

6

Note
  • Secondary elastic network interfaces (ENIs) cannot be bound to instances of this instance family when the instances are being created and can be bound to the instances after the instances are created. When you bind secondary ENIs to or unbind secondary ENIs from instances of the following instance types, make sure that the instances are in the Stopped state: ecs.t6-c1m1.large, ecs.t6-c1m2.large, ecs.t6-c1m4.large, ecs.t6-c2m1.large, and ecs.t6-c4m1.large.

  • You can go to the Instance Types Available for Each Region page to view the instance types available in each region.

  • For information about the specifications of the instances, see the Instance type specifications table in the "Overview of instance families" topic.

t5, burstable instance family

Features:

  • Provides a CPU performance baseline and the ability to burst above the baseline, which are governed by accrued CPU credits.

  • Balances compute, memory, and network resources.

  • Compute:

    • Offers multiple CPU-to-memory ratios.

    • Uses 2.5 GHz Intel® Xeon® processors.

    • Uses DDR4 memory.

  • Storage: supports only ultra disks and standard SSDs.

  • Network:

    • Supports IPv6.

    • Supports only VPCs.

  • Supported scenarios:

    • Web application servers

    • Lightweight applications and microservices

    • Development and testing environments

Instance types

Instance type

vCPU

Memory (GiB)

Average baseline CPU performance

CPU credits per hour

Max CPU credit balance

Network baseline bandwidth (Gbit/s)

Packet forwarding rate (pps)

NIC queues

ENIs

Private IPv4 addresses per ENI

ecs.t5-lc2m1.nano

1

0.5

20%

12

288

0.1

40,000

1

2

2

ecs.t5-lc1m1.small

1

1.0

20%

12

288

0.2

60,000

1

2

2

ecs.t5-lc1m2.small

1

2.0

20%

12

288

0.2

60,000

1

2

2

ecs.t5-lc1m2.large

2

4.0

20%

24

576

0.4

100,000

1

2

2

ecs.t5-lc1m4.large

2

8.0

20%

24

576

0.4

100,000

1

2

2

ecs.t5-c1m1.large

2

2.0

25%

30

720

0.5

100,000

1

2

2

ecs.t5-c1m2.large

2

4.0

25%

30

720

0.5

100,000

1

2

2

ecs.t5-c1m4.large

2

8.0

25%

30

720

0.5

100,000

1

2

2

ecs.t5-c1m1.xlarge

4

4.0

25%

60

1440

0.8

200,000

1

2

6

ecs.t5-c1m2.xlarge

4

8.0

25%

60

1440

0.8

200,000

1

2

6

ecs.t5-c1m4.xlarge

4

16.0

25%

60

1440

0.8

200,000

1

2

6

ecs.t5-c1m1.2xlarge

8

8.0

25%

120

2880

1.2

400,000

1

2

6

ecs.t5-c1m2.2xlarge

8

16.0

25%

120

2880

1.2

400,000

1

2

6

ecs.t5-c1m4.2xlarge

8

32.0

25%

120

2880

1.2

400,000

1

2

6

ecs.t5-c1m1.4xlarge

16

16.0

25%

240

5760

1.2

600,000

1

2

6

ecs.t5-c1m2.4xlarge

16

32.0

25%

240

5760

1.2

600,000

1

2

6

Note
  • Secondary ENIs cannot be bound to instances of this instance family when the instances are being created and can be bound to the instances after the instances are created. When you bind secondary ENIs to or unbind secondary ENIs from instances of the following instance types, make sure that the instances are in the Stopped state: ecs.t5-lc2m1.nano, ecs.t5-c1m1.large, ecs.t5-c1m2.large, ecs.t5-c1m4.large, ecs.t5-lc1m1.small, ecs.t5-lc1m2.large, ecs.t5-lc1m2.small, and ecs.t5-lc1m4.large.

  • You can go to the Instance Types Available for Each Region page to view the instance types available in each region.

  • For information about the specifications of the instance types, see the Instance type specifications table in the "Overview of instance families" topic.