zyxwvutsrqp

zyxwvutsrqp
zyxwvutsrqponm
Technical Editor:

zyxwvutsrqp
Marcin Paprzycki
Dept. of Computer Science and
Statistics
Univ. of Southern Mississippi
Southern Station 1506
Hattierburg, MS 39406-1506
m.paprryckiQurm.edu

~~~~~

High PerformanceCluster
~
High Performance Cluster
Computing

Because high-performance cluster comput-

ing is a relatively new area, few books success-
~~~

fully cover the topic. Rajkuniar Buyya provides

zyx
Reviewed hy Lalit A l . Patnnik, Zndinn Inrtitute of'Scimce

2s extensively deal with job and resource man-

agement systems, scheduling parallel jobs on
clusters, load sharing, parallel program sched-

zyxwvutsrqponm
Computing: a n authoritative overview of the field and its re]- uling, dyliainic load balancing, and mapping
evant state-of-the a r t research directions. Both and scheduling on heterogeneous systems.

zyxwv
Architecfuresand Systems,
vol. 1 volumes stem from Ius interaction with leading Part &'discusses representative cluster sys-
Programming and Applications, researchers, offering in-depth coverage for sci- tems such as Beowulf-class Linux clusters;
vol. 2
entists and engineers engaged in the research, COMPaS, a Pentiuin Pro PC-based SMI'
Rajkumar Buyya. ed.
development, and application of high-perfor- cluster; the BSP-Based Adaptive Parallel Pro-

zyxwvu
881 and 700 pages
154.00each mance computing systems. (Buyya's Web site, cessing System; and MARS, an adaptive par-
Printice Hail PTR \r?vw.dgs.monash.edn.ao/-rajkumar/cluster/ allel programming environment. A particu-
Upper Saddie River, New Jersey index.hmi1, offers a wealth of additional infor- larly interesting topic is a scalable and highly
1999 and 2000
0-13-013784-7and 0-13-013785-5 mation.) Graduate students should also find available cluster Web server. Chapter 36 pre-
these two books to be extremely useful, espe- sents key issues for buildinga Web server that
cially when exploring research topics. scales with increased computing reqnire-
mens, efficiently using cliister technology and
1: ARCHITECTURES
VOLUME AND allowing online maintenance while providing
SYSTEMS high availability of Web sewers.
Volume 1 contains 36 chapters divided into The in-depth coverage of so many topics
four parts. Each part features an introductory makes High l'eiforniance Chater Computing:
chapter that deals with key issues, making this A7.chitecrure.~nndSyrtcmr unique, and the bibli-
volume more accessible to readers with niin- ographies concluding each chapter are a great
iinal background. resource for recent publications and URLs.
Part I focuses on requirements and general However, the hook would have been niorc
issues, discussinghow to set up and administer a helpful had it provided advanced research-level
clnster and offering detailed information about problems a t the end of each chapter.
security policies and system monitoring. Part lI
covers the latest trends in networking, protocols, VOLUME
2: PROGRAMMING
AND
and UO. In particular, Chapter 9 offers detailed APPLICATIONS
information on high-speed networh, discussing Buyya separates the 29 chapters ofVolume
design issues, Fast Ethernet, high-performance 2 into thrcc parts, concluding each chapter
parallel interfaces, and asynchronous transfer with a n extensive bibliography, as he did in
modes. Chapter 14discwsses load balancingover Volume 1.
networks, and Chapter IS covers multiple-path Part I deals with programming environments
comnunicadon. Chapter 17 delves into an inter- and development tools. The first two chapters
esting trend in parallel computing-distributed concisely cover concepts related to parallel pro-
shared memory-discussing network perfor- gramming models, paradigms, languages, and
mance as well as design issues a t lengrli. environments. Chapters 3 and 4 discuss the
Part 111 covers process scheduling, load most popular message-passing environments,
sharing, and balancing. Chapters 20 through MI'I and PVM, and three software packages
-
86 IEEE Concurrency
 zyxwvutsr
zyxwvutsrq
that link thcsc two together-PLUS,

zyxwvutsrqpo
zyxwvutsr
I'ACX-MPI, and PVMPI. Part I also cov-
ers topics such as active objects, the (lis-

zyxwvuts
trilii~tcd sllarcd dara spstcm, and the
Lilirary for Pamllcl Systems, discussing
how to iniplcment distriliuted applica-
tiom 011 differcni arcliitcctitrcs ;ind hct-
crogencous coiiip~m" Chaptcr 9 goes
over the Tuple space paradigiti througli
Linda and Chapter 10 elegantly explains
how to dehug parallel codc. The last
chapter of Part I is particularly important,
'rhc editmid Iioard, staff, and I thank all thosc who reviewed manuscripts forlElB
Coiiriwrmy liist year. Your expertise and ciimtiiitinciit to high standards add immea-
suralily to Iioth the mngazine and to die field of concurrency. If wc oniittcd your
iiainc, we sincerely apologize; plcasc send 11s your complete contact information,
and wc'll do bciter ncxi time.

Gheith A. Abandah, Univ. ofYo7.r.rm

-J?.lz Brim, Editor-in-Chief

focusing on Web opcrating systems and Ishfaq Ahmad, Howg Kong Univ. of Srieiire and Technology
system infrastructurcs for efticicntly Stephen J. Allan, Utnh State Uniu.
developing wide-area applications. Gustavo Nonso, Imt. f i r Infbi7,intioi~.yrre?,re
Part I1 focuscs on Java in high-perfor- Reda Ammar, Univ. of Comecticiit
niance compnting, discussing diffcrent Jerzy Andrzej Rarchanski, Brock Uniu.
models that support distributed object- Twaii Basten, Eindhoven U~ziv,!f Technology
oriented coinpiitation. Chapter I 2 has Joachiin Baurnann, Uv;u.Smttgmt Breihuim~ixmam
excellent coverage of distributed-object Ahdelghani Bellaschia, Sleiirrnn Trcl,nologies
models such as RMI, CORRA, DCOM, Yosi Ben-Asher, Hnifn Univ.
and Voyager and discusses parallel pro- H. Ronald Berlack, ConzgumrtionMnnnpnent Inteozoriond
gramming models. Chapter 13 presents Prasanth Bhat, Univ. of Soathem C~lifoolnin
the bot topic of Web-based parallel com- Frank Ruschmann, Sicntem C o r p t e Terhaolagy
puting with Java, along with a case for
the J E T platforin. D. Cai, Utiiv. ofUlstertitJordanrro?un
Part 111presents newly developed and Philip Chan, IVoridri Inrt. of?'erhnology
newly applied algorithins and applica- Alok N. Choudhaty, Noithwertem Univ.
tions for a cluster environment. Chapter Paolo Ciancarini, Uniu. ofBolognn
17 is ofgreat interest bccause it discusses Andrea Clematis, Imt. per la Mntei,m;cn Applium
load balancing over lieterogeneoos work- Mark J. Clement, Brighmn Yomg Uniu.
stations, offering performance results. Carlo Combi, Uniuer.~itn'rlegliStiidi de UdIm
Part 111 also explains time management Crispin Cowan, Oregoti Grndmte Institnn
in parallel siinulation with sylichroniza- Thomas W. Crockett, NASA 1.nngIey Rcsenrch Center
tion protocols, algorithms, and hardware Mark Crovella, Uostnii Uziv.
system simulations with the network Michelle DiSanto, Uzio, dcgli Stirdi delSnnnio
enabled parallel simulator as a test bed. Weichang Du, Univ. ofNm Bnmwick
In addition, it details the implementation Tarek El-Ghazawi, George Mmon Uxiv.

zyxwvutsrq
ofseveral applicationssuch as Lazzy cval- Ophir Frieder, Illim~isI m . of Techmhgy

zyxwvutsrqponm
uation of parallcl ray tracing, image Kurt Gcihs, Univ. Frniikjwt
retrieval, databasc and knowledge base Apostolos Gerasoulis, Ketg-ers U i k

zyxwvutsrqpon
systems, climate ocean modeling, bio- Hans Michael Gerndt, CentrnLlnst. fbv Applied Mathe?imir\

zyxwvutsr
medical system modeling, and comlmta-
tional fluid dynamics simulation.
High l'eljbmniniirt. Ckrter Conipting:
Progunnzming a i d Ap/Jlirntiomis suitablc
for graduate shldents with a strong back-
ground in soliware for high-perfomiance
computers. It is ideal for a one-seniestcr
course with selected topics chosen kom
Harvey Glass, Univ. of'Soiith Plorida
Sanjay Goil, Pr$0in7nm Terhiialogy Groiq
Robert C. Goldstein, Uniu.of British Coli~mlL
Michael T. Goodrich, Jdvu Ilopkins U W .
Roberto Gorricri, Univ. N d q l i Smdi di Bolopm
Robert S. Gray, Daitlizaath Collep
Rachid Guerraoui, LSEUIEPFL CI<
Susanne E. Hambrusch, l'wdiir Univ.

zyxwvuts
Part 111. Almost all chapters include sec- Michi Henning, Piadin 7khmlogies
tions highlighting current research chal- David Holmes, Mucqi,mit Univ.
lenges; unfortunately, the hokwonld have Yennun Huang, Lucent Technologies
been more iotcresting and helpful had it Jean-Marie Hullot, INRk4
comewith a CD contaillingimpohlilt pro- Oscar H. Ibarra, Univ. of California, Sanrn Bnrlma
gramming examples. Also, unlike Voluirie Valerie Issamy, INRIANRISA
1 , a t times thc chapters seem disjointed. Xiaohua Jia, Cit, Uniu. ofHong Kong
Such dings happen when several authors Eric Jul, Utziv. ofCopenhqen
offer contributions on diverse topics, but Dennis Kafura, Virginin Foiyrerhnir Inst.
a minor reorganization of rhc cha ters Chaiidrikn Kamath, Lnwmn~-eLivn.r,rore National Lalomoly
could have overcome this prolilem. &
B William Phillip Kegelmeyer, Smdin Nntioiml Lnborntorie.?
_ _ _ _ . . ~ ~ - ~ ~

January-February 2000