Windows Azure configures the automatic scaling knob based on CPU utilization

One of the main benefits of the Windows http://www.aliyun.com/zixun/aggregation/13357.html ">azure platform" is the ability to quickly scale applications in the cloud to respond to fluctuations when needed. Before July last year, you had to write a custom script or use other tools, such as Wasabi or Metricshub, to enable automatic scaling. At the end of last June//build, Scott Guthrie announced that the automatic scaling function could be used on the local platform (also summarized in this article). This means that in the most common scenario, you no longer need to personally host Wasabi. Since then, it has become simpler to scale applications directly in the Windows Azure portal through configuration rules. Simply go to the Scale tab that applies to the cloud service or VM. The following example shows a knob that configures automatic scaling based on CPU utilization.

It was still in the preview state, and only basic metrics (CPU utilization and Azure queue Length) were supported. However, Windows Azure automatically scales to address the needs of most Azure customers. It's pretty straightforward. In fact, it's simple and intuitive, and you don't need to read an auto Flex tutorial to use it. We recommend that you consider this before browsing any other options or tools.

If your scenario requires more advanced functionality (such as other performance counters, time based constraints, composite rules, growth rates, custom metrics, or actions), Wasabi may still be a good choice during this period. Many of the features of Wasabi are in the planning table of the Windows Azure automatic scaling team, which will be added to future releases when appropriate. As for Wasabi, there is no plan to launch a future release. This is consistent with P&p's abandonment philosophy, where you can read more about it.

To help you understand the difference between the current version of the built-in auto-scaling feature and Wasabi, see the following table:

function

Windows
Azure Auto Flex

Since 2013
From June 26 onwards

Wasabi

Since 2013
From June 26 onwards

Integrated into Windows Azure portal

Is

Whether

Supported indicators

CPU
and queue Length

CPU, queue Length, and other Windows performance counters;

Easy to set

Extremely simple (built-in services in the platform and portal)

Medium (need to host a component)

Require dedicated storage for data points

No (this is a good thing!) ）

Yes (you can use Azure Blob storage or local file storage or customization)

Easy to configure

Extremely simple

Medium (need to configure storage account credentials and management certificates)

Impact on target applications

No

Need to enable the dependency counter to be captured in WAD

Support website

Is

Whether

Support Cloud Services (WEB role/worker role)

Is

Is

Support VM Roles

Is

Not tested

Custom Metrics

No, scheduled
API

Is

Custom Actions

Whether

Is

Cooling phase Support

Is

Is*

Timesheet based Scaling

No, but planned

Is

Composite rules

No, currently under consideration

Yes (supports expressions, nested functions, and aggregate functions)

Zoom Group

Whether

Is

Reasonable scaling decision History

Is

Is

Application restrictions

Whether

Is

Upgradeable

Automatically

Manually

Publish

Preview version

RTW

* Wasabi has 2 knobs: one knob is used to enable the cooling phase after any scaling operation is performed, and the other is used to optimize the cost of the billing boundary around hourly. Because Windows Azure now supports finer billing, it is recommended that you do not use the Wasabi optimization stabilizer.

During the keynote speech, Scott Guthrie listed Skype's example, one of the world's largest Internet services. Like most applications, Skype sees load fluctuations that cause unused capacity during off-peak hours. By migrating to Windows Azure and using automatic scaling, Skype achieves more than 40% of the cost savings (compared to running its own data center or not using automatic scaling).

This is where you can view other automated scaling case studies.

Regardless of the choice of automatic scaling function (recommended!) Or Wasabi, your application still needs to be designed for flexible design (for guidance, see the Wasabi documentation, the Cloud Development multi-tenant Application Guide (3rd edition), and CQRS
Most of the concepts in the Groceries Guide).

Auto-Scaling: Dynamic expansion/reduction based on real-world usage

One of the main advantages of Windows Azure is that applications can be dynamically scaled to respond to changing requirements. In the past, however, you had to manually change the size of your application or use other tools, such as WASABi or metricshub, to scale your application automatically. A year ago, we announced that automatic scaling would be built directly into Windows Azure. With the release of that day, automatic scaling has now been enabled for cloud services, virtual machines, and Web sites, and soon after that, support for a range of services, such as mobile services, has been launched.

With automatic scaling, you can configure Windows Azure to automatically scale your application dynamically (without manual intervention), so that you can achieve the desired performance and cost balance. When automatic scaling is configured, the number of instances that respond to the application load run is periodically adjusted. Currently, we support two different load metrics:

CPU usage percent
Storage queue Depth (for cloud services and virtual machines only)
In future updates, we will enable automatic scaling for more zoom metrics.

When to use automatic scaling

The following are good criteria for determining whether a service/application will benefit from the use of automatic scaling:

Services/applications can be scaled horizontally (for example, you can create multiple instance replicas)
Service/Application load changes over time
If your application meets these standards, you will want to take advantage of automatic scaling.

How to enable automatic scaling

To enable automatic scaling, simply navigate to the Scale tab in the Windows Azure management portal for the applications/services that you want to enable. On the Scale tab, open the automatic scaling settings for the CPU or queue (for cloud services and VMS) to enable automatic scaling. Then change the instance count and target CPU settings to configure the automatic scaling range to be maintained.

The following illustration shows how to enable automatic scaling on a Windows Azure Web site. I have configured the Web site to run with 1 to 5 VM instances. The exact number of uses will depend on the total CPU utilization of the VM (I have configured the scope below for using 40-70%). If the total CPU is more than 70%, Windows Azure automatically adds a new VM to the pool (I have configured up to 5 instances). If the total CPU is less than 40%, Windows
Azure will automatically shut down the VM to save costs:

When you turn on automatic scaling, you can return to the Scale tab at any time and select off to manually set the number of instances.

Using Automatic Scaling

With the update that day, you can now take a few minutes to get Windows Azure to automatically adjust the number of instances running on your application to keep service performance in good condition and at a lower cost.