Five ways to optimize the application of public cloud

Converting to a public cloud can be as simple as mobile apps and pay bills, but do you always spend more than you need? The following five public cloud application optimization technologies can help you save money and improve application performance.

Last month, the Friday night's storm caused the Amazon server to run out of service, as Netflix, Pinterest and Instagram crashed for services in the eastern United States. Netflix, in particular, was at its peak late Friday, so customers could not complain about the outage.

Similarly, Roudup is one of many genetic applications used to predict evolutionary relationships between genes, organs and biological functions. Rounduo uses intensive computing methods, so Harvard researchers have to synthesize the simple storage (S3), Flexible Computing Cloud (EC2), and flexible programming (EMR) services provided by Amazon Network Services (AWS). To optimize the application, the Roudup team reduced disk I/O, reduced the memory cache, and calculated the best number of instances required. As a result, Roudup saves 40% of its cost and ensures that it can adapt to future computing growth without impacting performance.

It may not be optimized and the application will function as Roundup. However, optimized applications can enhance application-related performance, such as practicality, disaster prevention, and, crucially, the cost of saving public cloud use. Here are 5 ways to optimize applications when converting to a public cloud:

1. Reorganization code, processing cloud http://www.aliyun.com/zixun/aggregation/15552.html "> Service provider Billing model

AWS does not just charge computing, storage and network broadband usage fees----each time you enter a memory to read and write, it collects fees. As a result, you may want to integrate read and write operations into your application, and you can merge both into a single operation at any time. In this way, you only need the cost of your own server, each subsequent read or write operation will not generate additional costs.

The overall effect of this cloud optimization technique depends on the pricing method of the public cloud service provider (CSP) that you are registering to use. However, no matter which CSP you register, code reorganization is an opportunity to improve application performance.

2. Optimize selected default entity cloud

When setting up EC2 instances, you can choose between different levels of computing, memory, and storage. In addition, spot examples are available on the EC2, and there are redundant examples that can be used at any time, and prices are far below the general price.

It is worthwhile for you to spend some time on your application trials to determine the best computing, memory, and storage levels needed. This will help you make sure that you don't overspend on performance or configuration, and it also helps to figure out if you should consider the spot instance (or equivalent obtained from another CSP).

3. The level of service required to balance the default entity cloud

All applications have their exclusive configuration----The service level of its general purpose and function. Customers-oriented E-commerce site, there and your internal staff portal a relatively different level of service. Estimating the cost of a public entity cloud, compared to the level of service required by a variety of applications, may help you optimize your own public cloud computing costs.

Review of the Netflix power outage of June 29. Given the nature of video streaming services, it may not be feasible to force it into Amazon's data center in another county. However, less intensive, and mission-critical services can be optimized to provide services beyond the available data centers, where possible, to avoid such outages.

4. Fine tuning Automatic extension principle

The ability to automatically adjust the number of server instances is everywhere, and this optimization provides a good opportunity. For example, once the CPU has an existing instance utilization rate of 80%, your self-tuning ratio ruler will produce a new instance, and if the average CPU utilization reaches 40%, the self adjustment ratio ruler will produce a new instance.

How do you know that 80% and 40% are the correct numbers? Why not 85% and 35%? With the latter rule, you will be able to generate fewer instances and lower costs.

In addition, each application algorithm, storage, bandwidth requirements vary. Therefore, your rules need to be based on the combination of the above three factors, not just CPU use. You may need to experiment with different combinations to find out what is appropriate for your public cloud application and what level of service you need.

5. Database Row optimization

Applications like Netflix have local limitations, that is, most of the time, customers only get access to some of the data that belongs to them. Netflix uses AWS regions and zones to run servers, providing services only to customers living around the data center.

This is thanks to database partitioning technology, which allows you to partition the rows in your database and store different partitions in different data centers in your database. The same technology applies to credit card processing, since zoning can be applied to local models, such as handling credit card holders ' information or transactions.

You do not have to store all the database rows in the database instance. If you can split the database nesting and store it in a different instance of the database partition, you can take advantage of the partition's usage model. This will require the number of instances, thereby reducing the cost of public cloud services.

When you convert the application to a public cloud, your application may run well without making any changes. However, if you take note of the CSP charging method and associate it with the calculation of application patterns, memory, storage, and network bandwidth usage, you can easily reduce the cost of the public cloud. Using code reengineering to optimize applications may improve application performance and prolong service life, while testing and fine-tuning the default instance and the self tuning ratio ruler may reduce CSP costs for you.

(Responsible editor: The good of the Legacy)