Content addressable parallel processors

4 downloads 0 Views 299KB Size Report
subject [ 1 1. The book starts with a short introductory chapter, “What is a Con- ... bilities made possible by simplifying assumptions in the structure of programs ...
988

PROCEEDINGS OF THE IEEE, VOL. 6 6 , NO. 8, AUGUST

1978

Book Reviews

m e ,follon$ng reviews were. selected,from those recentlv published in various IEEE TRANSACTIONS and GrouplSociety Newsletrers. They are reprinted here to make them conveniently available to the many readers who otherwise might not have readv access to them. Each review is followed by an idenr1jication of its original source.

Data Pmceasiag m 1980-1985: A Study of Pbtentid Limitations to Rogress-Silt Committee of SHARE Inc. (New York: John Wiley & Sons, 1976, 191 pp., $13.50). Reviewed by Ronald L. H W , Computer Science Corporation,Folk CRurdr, VA.

This treatiseonproblemsthat plague the “large, general-purpose business-oriented data processing systems” was authored by the Silt users oflarge data Committeeof SHARE Inc.,anorganizationof processing systems (predominantlythosesuppliedby IBM). The book‘s basic assumption is the continued development of all phases of computer technology along their current paths. This includes such predictions as “programming practices wiIl remain a craft,” and certain assumptions of world stability: no major war, and no major political, is based on a very coneconomic, or social revolution.Thebook servative attitudetowardsthefuture-a view thatincludesno new technologiesand ignores existingentities such as minis andmicros LSI technologies. and the effects of is contained in six chapters In the chapter The meat of the book on“environment,”the Silt Committeepredictstheneedfor additional security, research, and standardization; In the future, according to the committee, user demands will be signifcant. They wiIl require reliability, availability, consistency, understandability, simplicity, and forgiveness byfuturecomputerfacilitiesThis will come aboutby interfacing the installation control program with the vendor-supplied system control program.Presumably,eachinstallation will have its own ICP customizedandmodified to suit theinstallation,thereby removing the user one level further from the underlying system. Application, programs will capture data closer to the source as o p posed to having thedata arrive at the computational facility ready for mput on punched cards or magnetic tape. One of the implications of this “future” goal is the addition of more remote terminals and automated sensors attached to computersystemsAccording tothe will be Silt Committee, this ability to capture the data in real time complementedbytherequirement to maintain and process large data base on-line. In keeping with the book’s conservative ve i w, the major thrust of the hardware section is to increase the reliability and availability of the machine. To this end, the committee believes that fumware will be used morebymanufacturers,and less by users. Presumably, significant system software will be executed in microcode. It is usually advantageous to maintain continued vendor support by not modifying provided code. The“software”chapterdoespredicta“meansfortransferring responsibilityandaccountabilityfromoneprogrammertoanother.” This statement Seems tocontradicta previous one descriiingprogramming as a“craft.”TheSiltCommitteepredictsthatdata processing departmentsandthek management will begrantedequal statuswithothercorporatedepartmentsand managers. The committee argues that when a failure in any department can grind a company to a halt, that department must receive more attention. I believe this is evident already.

Much of the book Seems dragged out with opinions that project an IBM-like vision of the future: slow evolution rather thanradical change. Most ofthereferencesaretoarticlesofthelate 1960’s and early 1970’s. Thosereferring to technologiesofthemid-1970’sare dismissed as not being developed enough to be applicable by 1985. According to the Silt Committee, bubble memory technology is an example of ayet-to-badevelopedtechnology. And yet, slightly more than one year after the completion of this book, it is being marketed. In short, the authors give little credit t o developing technologies (both hardware and software), and overestimate the time needed for manufacturers to add new products to their computer lines. Placed ina1960timeframe, this book might have beenagood predictorofthe1970’sButinthe 1970’s timeframe,it seems to be little more than a record of what has been done and a suggestion of where we should try a little harder. Reprinted from Computer, April

1978.

EIectronics Designers’ Hmdbook-L. J. Giacoletto, Ed. (New York: McGraw-HiIl, 1977,2420 pp., $47.50). Reviewed by h n y L.

Harklau, Stillwater, MN.

This handbook can only be d e m i as a monumental undertaking. 2 pages ofreferencematerialwith30 It consists ofmorethan contributors, all of whom are ll known m thefield. Thehandbook covers all major areas mthe fiild ofelectronics, including materials,components,fiiters, devices, amplifiers, servo systems,modulationtechniques, receivers, transmitters,radar,and aspects of information theory,to name a few. The handbook has numerous tables and graphs to assist in design work. It has a good blend of theory and practical design implementation. Numerous examples are given to assist the designer in understanding the subject. For those engineers who may find any particular presentation too brief, there are numerous references given for more detail. If the handbook has a shortcoming, it is in the index, which could standmorecross-referencing.Forexample,thereare several tables with RC networks at various places in the text. The index lists the foIiowing under“RC”:activenetwork; ladder network;oscillators; To fiid tables on “RC transfer transferfunctions;twin-Tnetwork. impedances,” one must look under “impedances” in the index. Thehandbook is excellentfor engineers who have abroad range of responsibility or who frequently transfer from one electronics f i l d to another, such as those in research and development. It is also good for engineers who have been out of school for sometimeandfeel they need toupgradetheirtechnical capabilities. Engineers who prefer to work in specific areas may fmd a handbook covering their particular field of interest moreuseful. Thishandbookis highly recommended as areferencebookfor technical hiraries and those involved in electronicdesign.

3

Reprinted from IEEE Spectrum, April

1978.

Content Addressable ParaUel Rocessors-Caxton C. Foster (New Reinhold, 1976,233pp.,$11.95). Reviewed York:VanNostrand by Behrooz Parhami, Arya-Mehr University of Technology, Tehran, Iran. If theproliferationof survey articlesand specialized books in a particularareaof research isanyindicationofitsmaturity,then, after more than 20 years of research, the area of parallel and asso-

BOOK REVIEWS

ciativeprocessing is approaching that state.Thisbook “grew out of notes for a course in Content Addressable Memories that has been presented several times at the University of Massachusetts” and “attempts to gather into one place all of the information that exists as of late 1975.” Although this is the only book I know of which deals exclusively with associativeprocessing, others also have treatedthe subject [ 11. The book starts with a short introductory chapter, “What is a Content Addressable Parallel Processor and W h o Needs One?” Even though higher cost, the lack of urgent need, user resistance to change, and the general unfamiliarity of the concept have hampered their acceptance, Foster claims that “Content Addressable Parallel Processors will sweep all beforethemandbecomethe standardoftomorrow.” I share hisenthusiasm, but Ifeel that less optimistic viewsdeserve more mention than hegives them (asingle parenthetical statement of two such views at the bottom ofa page). After theintroduction of a simple m a w in Chapter 2, several definitions dealing withthe characterization of parallel executable algorithmsarepresented in Chapter 3 under the misleading title, “Some Theoretical Results.” Theauthor claims that “given aproposed method of solving aproblem(analgorithm), it is possible to determine if aproposedhardwareconfigurationcanpermit parallel execution of thisalgorithm” by using the characterizingdescriptors presented in this chapter. This appears, however, to be a direct result of the close correlationbetween the descriptors and hardware capabilities made possible by simplifying assumptions in the structure of programs under consideration. This is evident at the bottom of p. 48 in Chapter4where an exampledemonstratesthat, if computation results in one cell depend on the contents of other cells, then some form of intercell communicationis needed. to a discussion of algoAbout 45 percent of the book is devoted rithmsfor andapplications of associative processors in Chapters5 and 6. Although I have not checked the accuracy of all the algorithms presented, it is safe to say that these two chaptersarereasonably good,despitethefactthattheamountandorder of discussion of applications in Chapter 6 are not representative of their importance (one page and sixth position for air traffic control versus seven pages and third position for spelling error correction). However, the reader is warned of thevarying depth of treatment inadvance. Chapter7, “DistributedHardware,”dealswith mechanisms for multiple response resolution (selecting the fust responder) and counting the responders (finding theirexactnumber).Abriefmention (oneparagraph) is made of parallel input and output-atechnique which-may be very important in the design of low-costsystems in the near future and which was proposed as early as 1965 in the design of an associative fde processor [2]. The remaining three chapters deal with the design of severalassociative computer systems. Chapter 8 is devoted to a discussion of Staran,an associative computerbuiltandmarketedbyGoodyear Aerospace Corporation. This is followed by a relatively short chapter (16 pp.), “Other Content Addressable Machines,” which, as the author admits, is too brief to include a reasonable discussion of all interesting systems implemented or proposed. Finally, in Chapter 10, the author attempts to present a design for a semiconductor associative computer, including its intercell communication logic and control unit. The book contains a total of 16problems, 14 ofwhich are in Chapter 5 (“Algorithms”)-certainly too few if the book is intended for use as a course text. The problems are not very inspiring. Most of them are of the “Why?” and ‘Time this algorithm” varieties. The bibliography contains 81 entries, despite the fact that several hundred papers have been written on the uses and design of associative processors, as the author notes in the preface. The most recent references included are from mid-1974. The most serious shortcoming of the book,in my view, is theauthor’s total omission of alternatearchitecturesbytreatingonlybit-serial systems. Althoughadmittedlythesearethemostinterestingtypes, other classes are also of some significance, as suggested in a classifcationbythis reviewer (31, later also used by Yau andFung [4]. In ( 6 ) deserve at least some particular,block-orientedarchitectures[SI, discussionin abook on associativeprocessing which “has tried to summarize the fEld.” In summary, the book uses the same informal style as Foster’s book on computer architecture [7] and suffers from the same shortcomings: lack of precision, narrowness of view, and- absence of meaningful exercises. I do not recommendthisbook as a text; however, those with a serious interest in parallel and associative processing may fmd it a useful additionto their library, given the reasonable price.

989

REFERENCES

Thurber, Large Smle Computer Architecture-Parallel and Associative Processors, Hayden Book Co., 1976. R. H. Fuller, R. M. Bird, and R. M.Worthy, “Study of associative processing techniques,” Clearinghouse for FederalScientificand Technical Information, U.S. Department of Commerce, Document No. 0 6 2 1 5 1 6 , Aug. 1965. B. Parhami, “Associativememories and processors: An overview and selectedbibliography,” R o c . E E E , vol. 61, no. 6, June

I 1 1 K. J. I2I

131

1973. DD. 722-730. and H: S. 141 S.

S.’q,b

Fung, “Associative processor architecture-A survey, Computing Surveys, vol. 9, no. 1, Mar. 1977, pp. 3-27. 15 1 L. D. Healy, G . J. Lipovski, and K. L. Doty, “The architecture of acontext addressed segment-sequentialstorage,” AFIPS Conf. Proc., vol. 41, AFIPS Press, 1972, pp. 691-701. B. Parhami, “A highly parallel computing system for information retrieval,” AFIPS Conf. Roc., vol. 41, AFIPS Press, 1972, pp. 681-690.

I71

C. C. Foster, ComputerArchitecture. Co., 1970.

Reprinted from Computer, April

Van NostrandReinhold

1978.

Desisning the New City-A Systematic ApproPch-J. E. Gibson (New York: Wiley-Interscience, 1977). Reviewed by J. D. Palmer,Metmpolitan State College, Denver,CO 80204.

A better understanding of the nature, evolution, and future of cities may lead us to one solution to urban problems. According to Dr. J. E. Gibson, “If one plansa largenew freestandingcity, its geographic location will contributesomewhat toits success, but muchmore important will be the ability to attract and hold growth industry to provide an economic base.”Basic to the problems confronting cities now and in the foreseeable future is the present social trend of concentrating large numbers of poor in the core city. This concentration leads to negative social indicators related to quality of living factors such as crime, unemployment, welfare costs, etc. The American city is in trouble largely due to out migration of citizens and industry to save money. Dr. Gibson tells us that “‘aneconomic analysis’ is all that is [maybe] needed-nothingmetaphysical or psychological is required.” Marginal costs arecountedwhile all real costs incurred by the city are ignoredincludingsuch important costs as taxes, subsidizedservices(e.g., water and sewer), transportation systems, and the like. Afairset of issues andproblems is raised to establishthe decaying condition of most Americancities in order to set the stage to proposeasystemsanalyticalapproach to resolve the dilemma, i.e., existing cities are in rack and ruin, and new approaches must be taken to establish new cities. Gibson gives us a good insight into the problems and issues confronting the modem city. FromhereGibson moves to chapters. on suchtopicsas“Systems which Analysis Methodology” and “Goals forAmericanSociety” include verbalizing the systems approach and then taking the reader through the steps of a verbal application. The f i i sections develop the concept of a new city, give examples of earlier attempts to establish new cities, provide reasons for their failure, and end with details on how to plan for a new city. Given Americanandworldexperiencewith cities, new or old, we should really ask, why a new city at all when the old ones don’t really work?Gibson addresses thesequestionsfairlyandwithdeftness. He demonstrates that suburbs are totally dependent upon a core city for nearly everything from markets to services and adds the coup de grace by showing thattax revenues fromcentral citiesexceed tax expenditures of allkindseachyear. Havingestablished the requirement for cities of varyingsizes, Gibson proceeds to take us through the steps of a systemsanalysis of how to design a new city. This book should find great favor with the audience of social planners, sociologists,citizen groups,and the like, who do not possess wen-developed mathematical skills, as it provides a verbal mapfor applicationsofacomplexapproach to analysis,planning, and synthesis. Forthe accomplishedsystemsanalystthisbook will cause a great deal of scurrying to the bookshelf for the moremathematically inclined treatment. However, for both groups, this book, if carefully read,containsagreatmany insights intothe degree anddepthof human interaction required if any adjustment to living andworking patterns is to be developed. This insight extends to the teams which will approachthe problems as well asthose who will be affected. Gibson also provides the measure of interdependence and interaction between individuals and systems necessary for successful applications in all human endeavors. If there is a failing of any significance, it lies in the lack of including the seriousness of the energy problem and the consequences likely to