Five advantages in cloud computing development

Cloud computing is now booming, but the source of cloud computing is not something new in recent years. Early in the year, Oracle proposed that the receptacle Computer architecture was presented by Oracle's Receptacle Computer Inc., supported by Apple, Sun, IBM, and Netscpae. It can be said that cloud computing in the new bottle is actually a modified ' old wine '-netpc.

NetPC was supported by Intel, HP, Dell, and Compaq by Microsoft and Intel in April 1997. Whether it's an NC architecture or a NETPC architecture, it's the same thing, and NetPC is actually the Microsoft version of NC architecture (Sun and Oracle also have their own NCA); So another name called NetPC.

Basically, NC is to advocate that the PC function is not too strong, by connecting to the host to obtain the host in the hard disk space and computing power Support program in the host after the completion of the results returned to the PC.

But the theory of cloud computing and NetPC is highly different. From the thinking mode, NETPC focuses on the computing entity, while cloud computing focuses on services. From the commercial operation mode, at that time in the environment is mainly selling computers and operating systems; Cloud computers are now mainly to sell software and services.

Metrics on services provided by cloud computing

It is recommended that "MIPS" be used as a measure of the true value of a function in the virtual world, in order to avoid "the real world currency, because it can not unify the real physical meaning of the unit to measure the real value of goods, resulting in inflation and exploitation" and so on.

The development of cloud computing

The five main reasons are that distributed computing is bound to become more and more common and gradually develop into a mainstream computing model instead of a centralized mainframe computer:

1. The first reason for the distributed system now is that he has a better performance-price ratio than a centralized system. You don't have to pay hundreds of thousands of dollars for efficient computing.

2. Most applications are distributed in their own right. such as industrial enterprise application, management department and the scene is not in the same place.

3. High reliability. Redundancy is not only a necessary condition for biological evolution, but also information technology. Modern distributed system has highly fault-tolerant mechanism, and the control reactor mainly uses distributed to achieve high reliability.

4. Scalability. Buying a higher-performance mainframe, or buying a mainframe with the same performance, is much more expensive than adding several PCs.

5. High flexibility. Can be compatible with different hardware manufacturers, compatible with low configuration machines and peripherals to obtain high-performance computing.

Roughly calculated, the current personal computer CPU chip processing capacity of 200MIPS, is the implementation of 200M is 200 million times per second instruction, and recently Yahoo! company reported that they have achieved 10,000 nodes (node) is 10,000 PCs connected to the distributed system, The total processing capacity is 2,000,000mips, the fastest chip can not reach this speed, because the speed of the chip designed in a certain area is a limit, insurmountable. But the current world famous supercomputer so-called TOP500, achieves every second multi-trillion instruction execution, all is uses the distributed design, the world first IBM Bluegene supercomputer has adopted 32 racks, each frame deploys has 768 PowerPC440 CPU. By the way, this TOP500 basically uses the Linux operating system! Now the social and family-owned personal computer is the PC, only 30% of the computing power is used, or even lower, while the remaining 70% is actually idle, these idle computer resources and computing power only through distributed systems can be effectively used, which can greatly improve the computing power of a country, and the ability to compute is a measure of a country's national strength and scientific research capacity of indicators, which has not been fully recognized by the people.

The computing power of a country and region has now become an important strategic resource, no less than the importance of oil and other strategic goods. Cloud computing is the ability to connect ordinary servers or personal computers to get supercomputers, also called High-performance and high-availability computers, but at a lower cost. This is also an advanced project in the world. The cloud computing model can greatly improve the computer and commercial computing ability of our country, and make our country's economic competitiveness greatly improved. The United States and Europe have a wide range of distributed computing systems that mobilize and use these social computing capabilities for Human Genomics (genomics) research, astronomical problem studies, mathematical puzzle studies, and other scientific research issues. A study last year estimated that personal computer PCs in China were nearly 200 million.

According to the computer operating system guru Andrew S. Tanenbaum (AST) defines a distributed system: "A distributed system is a system that runs on multiple machines without shared memory, but in the eyes of the user as a computer". (from "Modern operating system", Machinery Industry Press, 1999 Chinese edition). Its goal is to make each user feel that a networked computer is a time-sharing system-like using a personal computer-rather than a collective of many computers, even a five-node distributed system should make users feel like they're using a 200,000-dollar mainframe machine, The only unusual feeling is that the processing speed is much higher and nothing else is different. For example, here's a simple example when a user in machine a uses a file that is installed in the user's directory on machine B and a user logs on to the directory of machine b using telnet command rlogin B, then this is not a real distributed system, because user a realizes the existence of another machine, A distributed system must do that when user a logs on to a directory without knowing whether he is on the local machine or on a directory on a remote machine, and that machine B is transparent for user A, which is the "transparency" requirements considered in the design of distributed systems. Other related issues include: problems with Distributed file systems, directory and file access mechanisms, and consistency issues, communication problems with distributed system processes, and so on. The current cloud computing strictly says it has not reached the semantics level of true distributed computing.