سایت CEB:ترجمه تخصصی رشته کامپیوتر - معرفی و دانلود کتب انگلیسی رشته کامپیوتر

امتیاز کاربران

ستاره غیر فعالستاره غیر فعالستاره غیر فعالستاره غیر فعالستاره غیر فعال
 

انتشارات: Taylor & Francis Group, LLC (Chapman & Hall/CRC)

اثر: Jacques Mohcine Bahi, Sylvain Contassot-Vivier, & Raphael Couturier

تعداد صفحه: 225

حجم: 935KB

جلد کتاب

توضیحات کتاب (خلاصه مقدمه):

Computer science is quite a young research area. However, it has already been subject to several major advance steps which, in general, have been closely linked to the technological progresses of the machine components. It can easily be assumed that the current evolution takes place at the level of the communication networks whose quality, either on the reliability or the efficiency levels, begins to be satisfying on large scales.

Beyond the practical interest of the data transfers, this implies a new vision of the computer tool in scientific computing. The last advances in large scale networks have permitted the emergence of clusters of clusters. Those meta-clusters are then commonly composed of heterogeneous machines linked together by a communication network generally not complete and whose links are also heterogeneous.

In fact, the addition of a machine in an existing parallel system, even if that machine is less efficient than the ones already in the system, increases the potential degree of parallelism of that system and thus enhances its performance. So, the heterogeneity of the machines does not represent any particular limitation in meta-clusters.

The efficient management of the heterogeneous communication links is still quite unexplored. Each hardware evolution often comes with a software evolution. Indeed, it is generally necessary to modify or extend the programming paradigms to fully exploit the new capabilities of the machines, the obvious goal always being a gain either in the quality of the results or in the time to obtain them, and if possible, in both of them. Hence, in the same way the parallel machines and local clusters have induced the developments of communication libraries in the programming languages, the emergence of meta-clusters implies an updating of the parallel programming schemes to take into account the specificities of those new computational systems.

The synchronous message passing model is completely satisfying with parallel machines and local clusters. But it is not the case with meta-clusters. In fact, even if distant communications are faster and faster, they are still far slower than the local ones. So, using synchronous communications in the programming of a meta-cluster is strongly penalizing due to the distant communications between the different sites.

The principle of asynchronous communication scheme is that it does not block the progress of the algorithm. So, during a communication, the sender does not wait for the reception of the message on the destination. Symmetrically, there is no explicit waiting period for message receptions on the receiver and the messages are managed as soon as they arrive. That allows the performance of an implicit and efficient overlapping of the communications by the computations.

Unfortunately, that kind of communication scheme is not usable in all the types of algorithms. However, it is fully adapted to iterative computations. Contrary to the direct methods, which give the solution of a problem in a fixed number of computations, the iterative algorithms proceed by successive enhancements of the approximation of the solution by repeating a same computational process an unknown number of times. When the successive approximations actually come closer to the solution, it is said that the iterative process converges.

In fact, under some conditions which are not too restrictive, the data dependencies between the different computational nodes are no more strictly necessary at each solicitation. In this way, they act more as a progression factor of the iterative process. Moreover, numerous scientific problems can be solved by using that kind of algorithm, especially in PDE (partial differential equations) and ODE (ordinary differential equations) problems. There are even some nonlinear problems, like the polynomial roots problem, which can only be solved by iterative algorithms. Finally, in some other cases such as linear problems, those methods require less memory than the direct ones.

The objective of this book is to provide theoretical and practical knowledge in parallel numerical algorithms, especially in the context of grid computing and with the specificity of asynchronism. It is written in a way that makes it useful to non-specialists who would like to familiarize themselves with the domain of grid computing and/or numerical computation as well as to researchers specifically working on those subjects.

آیکن نشانگر نوع فایل در سایت CEBدانلود


برای دانلود کتاب، ثبت درخواست ترجمه، و دریافت رایگان خبرنامه‌های سایت نیاز به ثبت نام دارید.

آیا از خواندن متون ترجمه‌ای بی سر و ته خسته شده‌اید؟ و بدنبال راهی برای بهره‌گیری از یک متن اصیل برای یادگیری هستید؟
آیا وقت کافی برای فهم و یا ترجمه‌ی مطلب خود در اختیار ندارید؟
از خواندن متنی که کارتان گیر آن است کلافه شده‌اید؟
کار را به ما بسپارید، خیالتان راحت!

هم اکنون! سايت CEB را به چند نفر از دوستان خود هم معرفی کنيد؛ با اين کار علاوه بر حمايت از ما، به بالا رفتن کيفيت خدمات و پايين ماندن تعرفه‌ی دانلود هم کمک کرده‌ايد.



dear author and publishers!
If you do not agree that your books be freely available through this site to Iranians - Those who are not subject to the Copy Right law - please contact us through your official email address so that we can identify you as the author or publisher of that books and remove all the books that you don't like to be accessible through this site. Note that only downloadable material can appear on this website. Also, note that this site is not the source of illegal publication of the books; We only gathered the books accessible via the Internet together and made these books more accessible to Iranians.


Valid XHTML 1.0 Transitional