Cloud rationing refers to the process of deploying and managing IT resources on a cloud infrastructure

The author of this article details a rationing performance test method that you can use to determine where the rationing performance lag occurs.

This article describes a rationing performance test method that you can use to determine which locations are lagging behind in cloud allocation performance. The purpose of this rationing performance test is to:

measures the total ration time from the user's point of view in an end-to-end manner. When multiple rationing is present, the trend of rationing time is judged. The entire allocation time is decomposed into several parts to determine which components and steps occupy the most performance overhead. When there are many rationing requests in the system, you get queued information at the component level to help identify bottlenecks.

Let's understand some of the basics of cloud rationing.

The basics of cloud rationing

Cloud rationing refers to the process of deploying and managing IT resources on a cloud infrastructure. It consists of three kinds of rationing:

virtual machine rationing includes instantiating one or more virtual machines that meet application hardware and software requirements. Most cloud vendors offer a range of VM templates with common software and hardware rationing. For example, Ibm®smartcloud Enterprise supports hundreds of types of VMS that use different processor, memory, and I/O performance options. Resource rationing: Mapping VMs to physical servers in the cloud and scheduling. Application rationing: Deploying specialized applications in a VM and mapping end user requests to application instances.

Customers can also ration their assets through Web portals and APIs, including virtual server instances, images, or persistent storage units.

This article focuses on virtual machine rationing and resource rationing, which are the two main features of the cloud. Let's look at the challenges of rationing performance.

The problem of rationing performance that can be circumvented

The rationing process is complex because virtual IT resources and network elements are unpredictable. Customers often experience problems with rationing performance, but it is difficult to determine which or what factors are causing them.

Some of the challenges that cloud customers face include:

Different cloud vendors use different rationing engines. Users must have a certain understanding of the rationing engine to effectively communicate with the cloud service provider on performance issues and identify the root cause of the problem. At run time, there may be unexpected interoperability performance issues that are not conducive to smooth rationing. Although some middleware components can be tested for performance and then integrated into the system, some performance problems occur only after interacting with other middleware, or with specific rationing to meet business requirements. In a large computing environment, such as a data center, the availability, load, and throughput of IT resources and networks can have an impact on performance. Rationing workflows can become complex and cause performance problems. For example, a common rationing engine provides services that allow specific middleware components, such as databases or application servers, to be provided. The actual underlying rationing operation is implemented through a specific engine script or component. Many operations include rationing services that can be integrated into different rationing workflows.

Let's look at rationing performance testing methods and analysis.

Rationing test methods and analysis processes

A team I worked with developed a test method that first describes a set of States and operations so that the entire rationing process can be defined independently of the specific rationing engine used. Figure 1 describes these states.

Figure 1. Define the rationing process through State and action instead of engine used

During the commit period (submission Period), the user submits a request and receives a response. If the rationing workflow is successfully invoked, the component's state will change from "accept" to "submitted". The status of the ration request changes to "New."

During resource reservation (Resource reservation Period), if all required components cannot be reserved, the complete solution cannot be allocated and the rationing process is aborted. To reserve a component, the state of the component changes from not Available to reserving. If the reservation succeeds, a reservation message is returned and the component's status is changed to Reserved.

During rationing (provision Period), as long as the rationing engine has completed all necessary steps to set up the component, a rationing message is sent and the component status becomes "provisioned". Once a component is provisioned, all of its properties can be obtained by running a custom log.

In this article, I'll do two tests: benchmarking and load Tests.

Benchmarking is necessary to achieve effective testing. I recommend testing some images with different types and sizes. In the benchmark, the total allocation time of the three phases is recorded and the timestamp is used for each state change. by calculating the cycle duration by request, you can probably understand which part takes up the longest time. The computing power and operating system version data of the virtual machines are also collected to detect rationing problems for specific image allocations.

Load tests simulate multiple allocations, which can cause longer rationing times. When you run a load test, you monitor component activity to find system bottlenecks. The timestamp for each component state change is recorded. Using the same type of image as a rationing target enables clearer comparisons and shows time trends. Observe the change in rationing duration with concurrent rationing, which helps you monitor component-level transactional behavior.

The team developed the test script based on the user's end-to-end workflow. The script sends a ration request and keeps track of the rationing status to determine if the rationing was successful, failed, or timed out. Performance testing tools, such as Rational®performance Tester, can be used to trigger rationing of workloads and capture client-side data, such as the image rationing and rationing cycles at the request level. Most rationing and management engines and tools provide native methods for resource and component operations.

As a result, when using such tools, customers can record which engines performed and which services/endpoints are invoked, and then calculate the ration cycle for each component level based on that data. Log analysis and report generation using Python and VBscript.