Книга: Distributed operating systems

1.2.1. Advantages of Distributed Systems over Centralized Systems

1.2.1. Advantages of Distributed Systems over Centralized Systems

The real driving force behind the trend toward decentralization is economics. A quarter of a century ago, computer pundit and gadfly Herb Grosch stated what later came to be known as Grosch's law: The computing power of a CPU is proportional to the square of its price. By paying twice as much, you could get four times the performance. This observation fit the mainframe technology of its time quite well, and led most organizations to buy the largest single machine they could afford.

With microprocessor technology, Grosch's law no longer holds. For a few hundred dollars you can get a CPU chip that can execute more instructions per second than one of the largest 1980s mainframes. If you are willing to pay twice as much, you get the same CPU, but running at a somewhat higher clock speed. 

As a result, the most cost-effective solution is frequently to harness a large number of cheap CPUs together in a system. Thus the leading reason for the trend toward distributed systems is that these systems potentially have a much better price/performance ratio than a single large centralized system would have. In effect, a distributed system gives more bang for the buck.

A slight variation on this theme is the observation that a collection of microprocessors cannot only give a better price/performance ratio than a single mainframe, but may yield an absolute performance that no mainframe can achieve at any price. For example, with current technology it is possible to build a system from 10,000 modern CPU chips, each of which runs at 50 MIPS (Millions of Instructions Per Second), for a total performance of 500,000 MIPS. For a single processor (i.e., CPU) to achieve this, it would have to execute an instruction in 0.002 nsec (2 picosec). No existing machine even comes close to this, and both theoretical and engineering considerations make it unlikely that any machine ever will. Theoretically, Einstein's theory of relativity dictates that nothing can travel faster than light, which can cover only 0.6 mm in 2 picosec. Practically, a computer of that speed fully contained in a 0.6-mm cube would generate so much heat that it would melt instantly. Thus whether the goal is normal performance at low cost or extremely high performance at greater cost, distributed systems have much to offer.

As an aside, some authors make a distinction between distributed systems, which are designed to allow many users to work together, and parallel systems, whose only goal is to achieve maximum speedup on a single problem, as our 500,000-MIPS machine might. We believe that this distinction is difficult to maintain because the design spectrum is really a continuum. We prefer to use the term "distributed system" in the broadest sense to denote any system in which multiple interconnected CPUs work together.

A next reason for building a distributed system is that some applications are inherently distributed. A supermarket chain might have many stores, each of which gets goods delivered locally (possibly from local farms), makes local sales, and makes local decisions about which vegetables are so old or rotten that they must be thrown out. It therefore makes sense to keep track of inventory at each store on a local computer rather than centrally at corporate headquarters. After all, most queries and updates will be done locally. Nevertheless, from time to time, top management may want to find out how many rutabagas it currently owns. One way to accomplish this goal is to make the complete system look like a single computer to the application programs, but implement decentrally, with one computer per store as we have described. This would then be a commercial distributed system.

Another inherently distributed system is what is often called computer-supported cooperative work, in which a group of people, located far from each other, are working together, for example, to produce a joint report. Given the long term trends in the computer industry, one can easily imagine a whole new area, computer-supported cooperative games, in which players at different locations play against each other in real time. One can imagine electronic hide-and-seek in a big multidimensional maze, and even electronic dogfights with each player using a local flight simulator to try to shoot down the other players, with each player's screen showing the view out of the player's plane, including other planes that fly within visual range.

Another potential advantage of a distributed system over a centralized system is higher reliability. By distributing the workload over many machines, a single chip failure will bring down at most one machine, leaving the rest intact. Ideally, if 5 percent of the machines are down at any moment, the system should be able to continue to work with a 5 percent loss in performance. For critical applications, such as control of nuclear reactors or aircraft, using a distributed system to achieve high reliability may be the dominant consideration.

Finally, incremental growth is also potentially a big plus. Often, a company will buy a mainframe with the intention of doing all its work on it. If the company prospers and the workload grows, at a certain point the mainframe will no longer be adequate. The only solutions are either to replace the mainframe with a larger one (if it exists) or to add a second mainframe. Both of these can wreak major havoc on the company's operations. In contrast, with a distributed system, it may be possible simply to add more processors to the system, thus allowing it to expand gradually as the need arises. These advantages are summarized in Fig. 1-1. 

Fig. 1-1. Advantages of distributed systems over centralized systems.

In the long term, the main driving force will be the existence of large numbers of personal computers and the need for people to work together and share information in a convenient way without being bothered by geography or the physical distribution of people, data, and machines.

Оглавление книги