The high availability of cloud computing is increasingly valued by the market

Recently, http://www.aliyun.com/zixun/aggregation/32714.html "> news media is more focused on cloud computing, so you will rarely hear about high availability." 5 years ago, the IT community was very focused on high availability and cluster development. But in the near term, things are different. In any case, high availability remains one of the key topics in the IT world, says Kai Dupke, senior product manager at SuSE.

It has a greater emphasis on cloud computing flexibility and low-cost advantages, so high availability topics seem to be rarely mentioned in cloud computing information.

In the past, UNIX and Linux platforms have traditionally assumed most of the high-availability (HA) jobs. The storage layer includes a RAID array, a network layer multi-layer network configuration, and an operating system that includes the HA feature, which ensures the proper operation of the application. Of course, there are some ha workloads at the application layer: The developer adds support to the cluster in the HA feature.

-

When enterprise customers migrate to more virtual infrastructures, such as private clouds or virtual data centers, HA is still in the infrastructure layer, not the application tier. The virtual layer may include some ha support, but it is still part of the infrastructure.

The situation in the public cloud is vastly different. In a public cloud, the requirements and expectations of the infrastructure layer are not as high as legacy systems. The infrastructure is just like a commodity, a penny a penny, so developers can only build HA functionality in their applications.

This is not to say that the public cloud is not good; the flexibility and cost of the public cloud are part of the reason. In addition, there may be practical logical problems in applying the HA principle to the public cloud. Japan has tasted the bitter fruit of failure in the early 2011, and the current technology cannot fully support the public cloud.

But HA is still a necessary part of the IT world, because not all IT departments need cloud computing.

First, migrating to cloud computing requires a lot of cost. Because today's cloud computing does not provide HA, this requires customers to rewrite their applications. Because cloud computing lacks a very important feature, customers have to spend money to improve their infrastructure.

The marketing department can make good use of this extra work. It's like selling a car without a steering wheel. "Because the steering wheel is not installed, so use your own steering wheel to ensure that no one can use your car," the statement aptly illustrates how some companies are selling cloud services.

In fact, the biggest obstacle to cloud computing is the lack of infrastructure support needed for business applications. The reason is complex, most applications are third-party applications that are not written by the company that uses the application.

To get HA functionality in cloud computing, you can ask these third-party companies to expose the application building and allow other companies to access it. But it's just an armchair study! This means that every third-party vendor will no longer work hard to develop its own program to incorporate HA into the application layer. This is why embedded ha (embedded ha in the operating system layer) is more acceptable to the majority of users.

I've only heard about ha in the private cloud because customers don't trust external service providers, not just ha, but very stringent service agreements.

The solution to this problem is to give customers a deeper understanding of cloud applications and to integrate HA into the cloud. How do you do that? You can assign workloads that can be allocated, and for a single workload (where almost every application has a single piece of data source), you can provide ha in the customer's operating system (using traditional installation mode in cloud services) or in the cloud infrastructure.

As a product manager, I can solve this problem in three ways:

* Provide ha in customer operating system;

* Provide remote control and management without installing software on each client;

* Provide infrastructure in the clouds ha;

In cloud computing, changing or restarting workloads works well until there is a cloud problem, when the customer starts asking for ha. Alternatively, customers can use cloud computing as a backup of the data center to obtain availability at the lowest system cost, but the cost of running is too high.

The latter requires automation of the system or a solution to the scalability of cloud computing. Some Japanese companies are trying to activate their cloud backups at once, but there will be downtime problems. This shows that the customer needs a structured solution and a strict service agreement.

However, configuring HA in cloud computing can apply ha to each cloud application at the right cost. This configuration is easy to control, eliminates the need for special construction and special consulting, gives customers the flexibility of cloud computing, and provides optimized clustering and workload.

We're not just talking about mission-critical systems. Configuring HA functionality on the company's Linux platform is more cost-effective than using expensive, proprietary HA tool kits.

It's a little quick to kill. But in today's business world, data and information flow are more important, and no one wants to waste time and money on a messaging system that crashes. We have been unable to tolerate the low availability of life.

In short, the server is no longer expendable. Business and it are inextricably linked, so there is no loss. HA is not as interesting as cloud computing, but it is always one of the focus of the IT world.

(Responsible editor: Lu Guang)