📄 node27.html
字号:
<html><!DOCTYPE HTML PUBLIC "-//W3O//DTD W3 HTML 2.0//EN">
<!Converted with LaTeX2HTML 95.1 (Fri Jan 20 1995) by Nikos Drakos (nikos@cbl.leeds.ac.uk), CBLU, University of Leeds >
<HEAD>
<TITLE>3.1 Defining Performance</TITLE>
</HEAD>
<BODY>
<meta name="description" value="3.1 Defining Performance">
<meta name="keywords" value="book">
<meta name="resource-type" value="document">
<meta name="distribution" value="global">
<P>
<BR> <HR><a href="msgs0.htm#2" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/tppmsgs/msgs0.htm#2"><img ALIGN=MIDDLE src="asm_color_tiny.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/asm_color_tiny.gif" alt="[DBPP]"></a> <A NAME=tex2html2177 HREF="node26.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node26.html"><IMG ALIGN=MIDDLE ALT="previous" SRC="previous_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/previous_motif.gif"></A> <A NAME=tex2html2185 HREF="node28.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node28.html"><IMG ALIGN=MIDDLE ALT="next" SRC="next_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/next_motif.gif"></A> <A NAME=tex2html2183 HREF="node26.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node26.html"><IMG ALIGN=MIDDLE ALT="up" SRC="up_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/up_motif.gif"></A> <A NAME=tex2html2187 HREF="node1.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node1.html"><IMG ALIGN=MIDDLE ALT="contents" SRC="contents_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/contents_motif.gif"></A> <A NAME=tex2html2188 HREF="node133.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node133.html"><IMG ALIGN=MIDDLE ALT="index" SRC="index_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/index_motif.gif"></A> <a href="msgs0.htm#3" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/tppmsgs/msgs0.htm#3"><img ALIGN=MIDDLE src="search_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/search_motif.gif" alt="[Search]"></a> <BR>
<B> Next:</B> <A NAME=tex2html2186 HREF="node28.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node28.html">3.2 Approaches to Performance Modeling</A>
<B>Up:</B> <A NAME=tex2html2184 HREF="node26.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node26.html">3 A Quantitative Basis for Design</A>
<B> Previous:</B> <A NAME=tex2html2178 HREF="node26.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node26.html">3 A Quantitative Basis for Design</A>
<BR><HR><P>
<H1><A NAME=SECTION02410000000000000000>3.1 Defining Performance</A></H1>
<P>
<A NAME=secmetrics> </A>
<P>
The task of the software engineer is to design and implement programs
that satisfy user requirements for correctness and performance.
<A NAME=3088> </A>
However, the ``performance'' of a parallel program is a complex and
multifaceted issue. We must consider, in addition to the execution
time and scalability of the computational kernels, the mechanisms by
which data are generated, stored, transmitted over networks, moved to
and from disk, and passed between different stages of a computation.
We must consider costs incurred at different phases of the software
life cycle, including design, implementation, execution, and
<A NAME=3089> </A>
maintenance. Hence, the metrics by which we measure performance
can be as diverse as execution time, parallel efficiency, memory
requirements, throughput, latency, input/output rates, network
throughput, design costs, implementation costs, verification costs,
potential for reuse, hardware requirements, hardware costs,
maintenance costs, portability, and scalability.
<P>
The relative importance of these diverse metrics will vary according
to the nature of the problem at hand. A specification may provide
hard numbers for some metrics, require that others be optimized, and
ignore yet others. For example, the design specification for an
operational weather forecasting system may specify maximum execution
time (``the forecast must complete within four hours''), hardware
costs, and implementation costs, and require that the fidelity of the
model be maximized within these constraints. In addition, reliability
is of particularly high importance, as may be scalability to future
generations of computers.
<P>
In contrast, a group of engineers developing a parallel database
search program for their own occasional use may be happy with
anything that runs faster than an existing sequential program but
may be tightly constrained as to how long they can spend on
implementation. Here, scalability is less critical, but the code
should adapt easily to changes in both the database system and
computer technologies.
<P>
As a third example, consider an image-processing pipeline consisting
of several concurrent stages, each performing a different
transformation on a stream of images. Here, one may be concerned not
with the total time required to process a certain number of images but
rather with the number of images that can be processed per second
(<i> throughput
</i>) or the time that it takes a single image to
<A NAME=3091> </A>
pass through the pipeline (<i> latency
</i>). Throughput would be
<A NAME=3093> </A>
important in a video compression application, while latency would be
important if the program formed part of a sensor system that must
react in real time to events detected in an image stream.
<P>
In other situations, the ratio of execution time to system cost may be
important. For example, consider a bank that spends two hours every
night on its overloaded mainframe computer running an analysis program
that searches for fraudulent transactions. A version that runs in six
hours on a parallel computer at one-twentieth the cost is
significantly more cost effective even though the total execution
time is greater.
<P>
Much of the material in the rest of this chapter is concerned with the
modeling and measurement of just two aspects of algorithm performance:
execution time and parallel scalability. We focus on these issues
because they are frequently among the more problematic aspects of
parallel program design and because they are the most easily
formalized in mathematical models. However, these topics must be
examined in the context of a broader design process that also
addresses the other issues listed in this section.
<P>
<BR> <HR><a href="msgs0.htm#2" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/tppmsgs/msgs0.htm#2"><img ALIGN=MIDDLE src="asm_color_tiny.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/asm_color_tiny.gif" alt="[DBPP]"></a> <A NAME=tex2html2177 HREF="node26.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node26.html"><IMG ALIGN=MIDDLE ALT="previous" SRC="previous_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/previous_motif.gif"></A> <A NAME=tex2html2185 HREF="node28.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node28.html"><IMG ALIGN=MIDDLE ALT="next" SRC="next_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/next_motif.gif"></A> <A NAME=tex2html2183 HREF="node26.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node26.html"><IMG ALIGN=MIDDLE ALT="up" SRC="up_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/up_motif.gif"></A> <A NAME=tex2html2187 HREF="node1.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node1.html"><IMG ALIGN=MIDDLE ALT="contents" SRC="contents_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/contents_motif.gif"></A> <A NAME=tex2html2188 HREF="node133.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node133.html"><IMG ALIGN=MIDDLE ALT="index" SRC="index_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/index_motif.gif"></A> <a href="msgs0.htm#3" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/tppmsgs/msgs0.htm#3"><img ALIGN=MIDDLE src="search_motif.gif" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/search_motif.gif" alt="[Search]"></a> <BR>
<B> Next:</B> <A NAME=tex2html2186 HREF="node28.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node28.html">3.2 Approaches to Performance Modeling</A>
<B>Up:</B> <A NAME=tex2html2184 HREF="node26.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node26.html">3 A Quantitative Basis for Design</A>
<B> Previous:</B> <A NAME=tex2html2178 HREF="node26.html" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/node26.html">3 A Quantitative Basis for Design</A>
<BR><HR><P>
<P><ADDRESS>
<I>© Copyright 1995 by <A href="msgs0.htm#6" tppabs="http://www.dit.hcmut.edu.vn/books/system/par_anl/tppmsgs/msgs0.htm#6">Ian Foster</a></I>
</ADDRESS>
</BODY>
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -