Subvert the traditional work style! Cloud computing will keep us away from hardware

If you're working in a system administrator or IT manager, then cloud computing is likely to fundamentally change the way you work in a short time. The Vice president of Sun Engineering, Hal Stern, said in a speech at the Securities and Financial Markets Association (SIFMA) technical Management conference.

"The development of cloud computing is getting us out of the hardware world," Stern said.

What he means is that the system administrator's work from then on will become a monitoring network, without much attention to the hardware component. "We still need a system administrator, but they don't have to run around in the data center with a wrench," he said. Instead, they need to use telemetry technology and tools to evaluate the traffic, security, and performance of the network. ”

What are the characteristics of cloud services?

According to Gartner's survey, the cloud has five properties. It is service based; it is scalable and resilient, adding and removing underlying schemas as needed; it can use the shared underlying architecture to achieve economies of scale; it can measure performance, pay for user usage, and, of course, it uses Internet technology.

Some companies do not want to share the underlying architecture with others, so they create private cloud. The rest of those who are more concerned with price factors and are willing to share the underlying architecture with other companies can choose the public cloud (cloud).

"In Sun's view, there will be many types of clouds in the future." There will be private clouds and public clouds, and there will be new kinds of clouds between them, although the overall distribution of the cloud is still small between the two sides, like the barbell shape, "Stern said." If 51cto.com readers are interested in cloud services, see the following article, "cloud services competition to increase cloud computing standards to be unified" "2012 Enterprise cloud computing and cloud services spending will reach 42 billion dollars."

Address the tension between developers and policymakers

Cloud computing can solve a headache problem for common IT managers and some special system administrators, Stern said. "Cloud computing can help address the tension between developers and policymakers. This tension is like the story of the Garden of Eden. If you don't want the developer to take a bite, why put the apple there, "he joked, comparing it decision-makers to God."

He said: "IT managers ask: Why do users have so many needs?" How do I decide what they're doing? Why do they need so many versions? ”

Cloud computing will make all of this easier. Applications can be better monitored and deployed, depending on the specific needs of the enterprise.

"For startups, the cloud allows them to spend money on wages, project development and beer-without having to invest in the underlying structure," Stern said.

Using cloud services is easier to monitor, but Stern warns that paying on demand does not always mean saving money. It's like driving, if you drive every day, you should buy one, but if you don't drive a lot, it's more appropriate to rent a car, like 51cto.com's article, "CIO wasted 10 sins: How to use the budget to the edge?" "It is not important to spend much money, it is important to spend it on the blade."

Sometimes, business managers can use the cloud to avoid using it power. Stern told the story of Derek Gottfrid, the New York Times data architect, who used Amazon's AWS and Hadoop parallel Data processing architecture to move the 70 newspaper content into the Timesmachine archive, bypassing the IT department alone to accomplish this feat.

The reasonable data processing capacity of cloud computing today is about 810,000 PNG images and 405,000 JavaScript files, of course, but it's not the best choice if you have to deal with so much data every day.

Future Cloud Architecture

Stern says the IT department needs to change its working mindset to manage the underlying architecture that will be composed of many pieces of cloud in the future. He said that the underlying architecture should now be built through URLs and APIs instead of reading and writing commands, as 51cto.com's previous article, "Tsinghua Zheng People: cloud computing has to start with infrastructure."

Cloud applications are more rapidly built than traditional applications. But in response, the system administrator needs to give up the equation of their favorite measurement reliability.

"In the 90 's, we relied on mtbf/(mtbf+mttr) to measure reliability, MTBF was the average time to work without downtime, MTTR was the average recovery time," Stern said.

This equation, he says, usually produces a percentage number that has many 9. As an administrator, he said, looking at the equation, they would be inexplicably worried that there would be a sudden outage, even if it only happened once a year.

"So we try to reduce the average recovery time by constantly buying hardware." We bought raid, SAN and various clusters, "he said." "But software and deployment also affect reliability. ”

Now the system administrator needs to track a series of KPIs (key configured indicator, critical performance metrics) to measure service performance. "The system administrator needs to measure predictability (predictability), integrity (integrity), productivity (productivity) and efficiency (efficiency)," says Stern, who refers to the four factors as pipe.

He added that future system administrators would measure the efficiency of data centers based on data throughput per dollar and per watt of energy.

The system administrator will not track hardware performance in the cloud. "Even in the private cloud, we may not see the underlying hardware, let alone the hardware of the public cloud," he said. ”

The system administrator will no longer have to worry about replacing the hardware. "If you have 10,000 servers, there's always some problems you can't avoid," Stern said. " ”

He said that system administrators will no longer need to focus on the average server recovery time, but focus on how to quickly identify and solve problems. They need to clear the old instance of the application in time and add more instances when needed. In the cloud, slow reaction time equals failure.

He added that while job-function changes are not easy, there will be a large number of cloud service providers vying to help you.

"Keep in mind that the new role of the system administrator will focus on how to identify and solve problems rather than make a few more whenever on reliability," he concludes.