• ベストアンサー
  • すぐに回答を!

翻訳をお願いしたいです。コンピューター関係の書物の文章です。

翻訳をお願いしたいです。コンピューター関係の書物の文章です。 Computation refers to the processing that can be done in a unit time. Using processors in parallel can increase computational power (as can waiting for a year or two― a reasonable rule of thumb is that computational speed doubles every 18 months!). We measure computation in millions of instructions per second (MIPS). We assume here that every instruction takes the same time to execute, which is more or less true for reduced-instruction-set (RISC) processors. In mid-1996, even low-end Intel 80486-class microprocessors processed about 30 million instructions per second. At the high end, Alpha processors from Digital Equipment processed more than 500 million instructions per second. With operating-system overheads, however, only about 80% of this rate is available to applications.

noname#111799
noname#111799

共感・応援の気持ちを伝えよう!

  • 英語
  • 回答数2
  • ありがとう数2

質問者が選んだベストアンサー

  • ベストアンサー
  • 回答No.2
  • SPS700
  • ベストアンサー率46% (15293/33009)

 「コンピュテーションは、単位時間に行われる運算を指す。プロサッサーを、並列に使えば、1年か2年の待ち時間分(演算速度は18ヶ月ごとに2倍になると言われているので、妥当な目安である)演算能力を増すことが出来る。  演算は、1秒間に百万回(MIPS)の単位で計算する。ここではどの演算も結果が出るまで同じ時間がかかると見なすが、リデュースト・インストラクション・セット(RISC)のプロセッサーの場合、大体において正当と見てよい。  1996年半ばでは、Intel 80486クラスの低性能マイクロプロセッサーでさえ、30 MIPS の演算が出来た。  Digital Equipment 社製、高性能の Alpha プロッセッサーは、500 MIPS 以上のプロセスが可能であった。  しかしながら、諸経費の関係から、実際に使われているのは、この数字の約80%に過ぎない。

共感・感謝の気持ちを伝えよう!

その他の回答 (1)

  • 回答No.1
  • bakansky
  • ベストアンサー率48% (3503/7246)

コンピュテーションというのは、単位時間内になされるプロセスのことである。 複数のプロセッサーを同時に使えば計算能力を増大させることができる(1、2年、大雑把なところ18ヶ月毎に計算速度は倍になるともいえる)。 性能は1秒間に何百万の計算を処理できるか(MIPS)で評価される。 どの命令も実行時間が同一であると仮定すると、これは多かれ少なかれ縮小命令セット(RISC)プロセッサーについては当てはまる。 1996年中頃には、ローエンドの Intel 80486 というマイクロプロセッサーでは毎秒3千万の命令を処理できた。 ハイエンドのものになると、デジタル・イクイップメントの Alpha というプロセッサーは毎秒5億回の命令処理を行うことができる。 しかしながら、オペレーティング・システムがオーバーヘッドを伴うために、その能力の80%しかアプリケーションに回されない。

共感・感謝の気持ちを伝えよう!

関連するQ&A

  • 翻訳をお願いしたいです。コンピューター関係の書物の文章です。

    翻訳をお願いしたいです。コンピューター関係の書物の文章です。 Consider a building that has two floors, with an escalator to carry people from one floor to the other. Ignoring queuing delays, the response time for a passenger is the mean time taken by the escalator to ascend or descend one floor. The throughput (bandwidth) is the mean number of passengers that can be loaded or per second. Suppose that an average of five people step on the escalator in one second, and that the escalator takes an average of 10 seconds to go up one floor. The response time for a passenger, therefore, is 10 seconds, and the throughput of the escalator is 5 passengers/second. Thus, the degree of parallelism, which is the mean number of passengers carried simultaneously, is 5*10=50. To see this, mark a passenger with a daub of red paint as she steps on the escalator. In the ten seconds that she takes to reach the top, we expect that fifty more passengers boarded the escalator. Thus, when she steps off, the escalator carries an average of fifty passengers, which is its degree of parallelism.

  • 翻訳をお願いしたいです。コンピューター関係の書物の文章です。

    翻訳をお願いしたいです。コンピューター関係の書物の文章です。 we measure space in kilobytes(KB) or megabytes(MB), and bandwidth in kilobit/second(Kbps) or megabits/second(Mbps). Unfortunately, while a kilobit/second means 1000bits/second, a kilobyte is not 1000bytes, but 1021bytes. Similarly,1Mbps is 1,000,000bps, but 1megabyte is 1,048,576bytes. For back-of-envelope calculations, we can assume that an 8-Mbps link carries 1megabyte in 1second. However, for more precise calculations, a careful engineer must make the necessary conversions. In this book, we will always do so.

  • 卒論に使用するため、翻訳をお願いしたいです。コンピューター関係の書物の

    卒論に使用するため、翻訳をお願いしたいです。コンピューター関係の書物の文章です。 Consider a building that has two floors, with an escalator to carry people from one floor to the other. Ignoring queuing delays, the response time for a passenger is the mean time taken by the escalator to ascend or descend one floor. The throughput (bandwidth) is the mean number of passengers that can be loaded or per second. Suppose that an average of five people step on the escalator in one second, and that the escalator takes an average of 10 seconds to go up one floor. The response time for a passenger, therefore, is 10 seconds, and the throughput of the escalator is 5 passengers/second. Thus, the degree of parallelism, which is the mean number of passengers carried simultaneously, is 5*10=50. To see this, mark a passenger with a daub of red paint as she steps on the escalator. In the ten seconds that she takes to reach the top, we expect that fifty more passengers boarded the escalator. Thus, when she steps off, the escalator carries an average of fifty passengers, which is its degree of parallelism.

  • 翻訳をお願いしたいです。コンピューター関係の書物の文章です。

    翻訳をお願いしたいです。コンピューター関係の書物の文章です。 By explicitly identifying performance metrics and resource constraints, a system designer ensures that the design space is well defined, the solution is feasible, and the design is efficient. She can then trade unconstrained resources for constrained ones to maximize the design's utility at the least cost. Continuing with our example, a system designer might use the PC's surplus computational power to compress data as much as possible, to best exploit the limited capacity of the transmission link. A well-designed system maximizes achievable performance while still satisfying the resource constraints.

  • 翻訳をお願いしたいです。コンピューター関係の書物の文章です。

    翻訳をお願いしたいです。コンピューター関係の書物の文章です。 Nevertheless, it is still, possible to identify some principles of good design that have withstood the test of time and are applicable in a variety of situations. In Section6.2, we will study some common resources, so that the reader can get some intuition in identifying them in real systems. We will then build up, in Section6.3, a set of tool to help us trade freely available (unconstrained) resources for scarce (constrained) ones. Properly applied, these tools allow us to match the design to the constraints at hand. Finally, in Section6.4, we will outline a methodology for performance analysis and tuning. This methodology helps pinpoint problems in a design and build a more efficient and robust system.

  • 翻訳をお願いしたいです。コンピューター関係の書物の文章です。

    翻訳をお願いしたいです。コンピューター関係の書物の文章です。 If we could quantify and control every aspect of a system, then system design would be a relatively simple matter. Unfortunately there are several practical reasons why system design is both an art and a science. First, although we can quantitatively measure some aspects of system performance, such as throughput or response time, we cannot measure others, such as simplicity, scalability, modularity, and elegance. Yet a designer must make a series of trade- offs among these intangible quantities, appealing as much to good sense and personal choice as performance measurements. Second, rapid technological change can make constraint assumptions obsolete. A designer must not only meet the current set of design constraints, but also anticipate how future changes in technology might affect the design. The future is hard to predict, and a designer must appeal to instinct and intuition to make a design "future-proof." Third, market conditions may dictate that design requirements change when part of the design is already complete. Finally, international standards, which themselves change over time, may impose irksome and arbitrary constraints. These factors imply that, in real life, a designer is usually confronted with a complex, underspecified, multifactor optimization problem. In the face of these uncertainties, prescribing the one true path to system design is impossible.

  • 翻訳をお願いしたいです。コンピューター関係の書物の文章です。

    翻訳をお願いしたいです。コンピューター関係の書物の文章です。 Besides constraints on these five resources, two major constraints are purely social :standards and market requirements. ・Standard force the design to conform to some widely agreed-upon requirements that may or may not be appropriate for that particular design. Though standards are essential to allow interoperability and can reduce costs by promoting competition, poorly designed standards severely restrict the design space. Worse yet, an underspecified standard may lead to faulty implementations that are not even interoperable. A good example of a bad standard is the Q.2931 standard for signaling in ATM networks. The standard is unnecessarily complex, and therefore hard to implement and hard to modify. Yet, to succeed in the marketplace, switch manufacturers have to obey it, thus slowing the rate at which ATM networks spread. ・Market requirements, like poorly designed standards, are unfortunate realities. A company may decide that products being designed must be backward-compatible with a previous product, or that they must use a particular operating system, or that they must support some set of standards. These essentially political decisions set the framework for a design. Market requirements may also determine the amount of human effort that can be expended on a design project. It is remarkable how much time is spent on building graphical user interfaces, sometimes at the expense of the system underlying the interface!

  • 翻訳をお願いしたいです。コンピューター関係の書物の文章です。

    翻訳をお願いしたいです。コンピューター関係の書物の文章です。 Standard force the design to conform to some widely agreed-upon requirements that may or may not be appropriate for that particular design. Though standards are essential to allow interoperability and can reduce costs by promoting competition, poorly designed standards severely restrict the design space. Worse yet, an underspecified standard may lead to faulty implementations that are not even interoperable. A good example of a bad standard is the Q.2931 standard for signaling in ATM networks. The standard is unnecessarily complex, and therefore hard to implement and hard to modify. Yet, to succeed in the marketplace, switch manufacturers have to obey it, thus slowing the rate at which ATM networks spread.

  • 翻訳をお願いしたいです。コンピューター関係の書物の文章です。翻訳サイト

    翻訳をお願いしたいです。コンピューター関係の書物の文章です。翻訳サイトのコピペはご遠慮ください。 Consider an airline reservation system, where agents from any part of the world can make a reservation for seats on any flight on any airline. One design for this system is to send all reservation requests to a single central computer. This design is simple, but has two problems. First, if the central computer crashes, every agent is affected. Second, as the number of agents increases, we need to expand the capacity of the central computer. However, the number of reservation requests, particularly during peak travel periods, may increase beyond the capacity of the largest computer that we can build or buy. Then, the response time suffers, and system performance degrades. We can solve this problem by replacing the central computer with a set of regional reservation center that coordinate among themselves to maintain a consistent view of the system state(such as whether a night is full or not). Then, as the number of reservation requests increases, we can just add another reservation center. We must, however, pay for this with a communication overhead for coordination, and a complex network to interconnect the regional reservation centers.

  • 翻訳をお願いしたいです。コンピューター関係の書物の文章です。翻訳サイト

    翻訳をお願いしたいです。コンピューター関係の書物の文章です。翻訳サイトのコピペはご遠慮ください。 The last constraint, scaling, is different from the others in that it constrains, not resources, but design elements. A centralized design, that is, a design that incorporates a single point of storage or control, is limited to the storage or processing capacity of the central point. Thus, an arbitrarily scalable design must minimize its use of centralized elements. This forces such designs to use distributed algorithms for control and coordination. Of course, a distributed system has its own problems, such as communication overhead and complex interconnection. Scaling is hard to measure, and arguments about whether a design can scale verge on the religious. Nevertheless, history shows that a system must scale well if it is to succeed.