This web log contains editors’ comments and observations on papers published in the IBM Journal of Research and Development and the IBM Systems Journal. Readers are invited to contribute comments on these postings or on any paper ever published in the IBM Journals.

Papers on Hardware Design and Implementation

The section begins with A. S. Hoagland’s 1961 paper “A high track-density servo-access system for magnetic recording disk storage.” Hoagland describes special techniques to maximize track density in random-access mass storage. In particular, he discusses a servo-access concept that provides self-adjusting head positioning. The paper also discusses in detail a novel method for precisely positioning the recording head and providing track location and registration at 154 tracks per inch. The IBM 3330 Data Storage Unit, announced in 1970, was the first production disk file to incorporate a track-following feedback control system.

Papers of more recent vintage include the 2006 paper “Using advanced compiler technology to exploit the performance of the Sony Cell Broadband Engine™ Architecture” by A. E. Eichenberger and colleagues. The Cell Broadband Engine microprocessor was jointly developed by Sony, Toshiba, and IBM and was designed to bridge the gap between conventional desktop processors and more specialized high-performance processors. Additionally, the processor has potential for use in digital distribution and entertainment systems—as well as for scientific imaging systems and physical simulations. Because game applications often feature highly parallel code for physical simulations and graphical realism, game developers require fast response times and a full-featured programming environment. The Cell Broadband Engine Architecture targets such applications by using a 64-bit multithreaded PowerPC^® processor element (PPE) with two levels of globally coherent cache and eight synergistic processor elements (SPEs). This paper also addresses an ongoing critical technical challenge in which the growth in processor complexity drives the need for sophisticated compiler technology.

Other papers are also highlighted in this section on “Hardware Design and Implementation” and include such topics as inductance calculations in a complex integrated circuit environment; engineering designs for a disk storage facility with data modules; developments in magnetic storage, tape subsystems, electronic packaging, solid-state memory, and semiconductor logic technology; tools for logic synthesis and characterizing long on-chip interconnections; processors for the IBM RISC System/6000^® and IBM System/390^®; an overview of the POWER4™ system microarchitecture; and “Millipede” (a data storage concept that is based on the atomic force microscope).

Some of the papers are interesting from a historical standpoint because they describe trends such as the profound increase in density and speed in memory technology—as well as the decrease in power and cost—that has taken place over a period of 25 years. For example, some of the papers review the ways in which processor memory technology has evolved from ferrite cores and thin magnetic films in the 1950s and 1960s, to semiconductor memories in the late 1960s through the 1970s, along with the various innovations that made this revolution possible. Technology has certainly come a long way since some of the earliest memories used in electronic calculating machines, such as the mercury sonic delay line developed in the late 1940s. This serially addressed memory required several milliseconds to read or write information.

Since delivery of the first movable-head disk drive in 1957, tape and disk devices in many configurations have been the principal means for storage of the large volumes of data required by data processing systems. IBM has developed several magnetic storage products that were the first of their kind and provided a significant new functional capability for computing systems. The evolution of each of these products can be identified with technological innovations that go beyond basic magnetic recording technology and include innovations that range from vacuum-column tape drives with high-torque, low-inertia motors to flexible diskettes and cartridge library storage facilities.


Nov 19 2007, 03:49:49 PM EST Permalink

Papers on Device Materials and Processes

The first paper describes a solution to an early packaging concern, i.e., an inability to increase the number of transistors mounted on a module using the available technology. In 1969, Miller proposed reducing the area of substrate lands and chip contact terminals and limiting solder to a small pad area. Upon heating, surface tension in the molten solder pad and in the land solder would support the device until the joint solidified. Variations of this approach remain in use today.

In the 1960s, aluminum was used to interconnect devices on a chip. The goal for future circuits was to increase the circuit density with many smaller devices. An increase in performance would require higher current densities. Ames et al. (1970) reported that aluminum films subjected to high current densities and elevated temperatures could fail by a process known as “electromigration,” which ultimately induces interconnection failures. Using similar conditions, they observed that the addition of 4% copper to the aluminum significantly retarded failure in the films. Thereafter, aluminum with 4% copper was used to interconnect components in integrated circuits. In the 1990s, aluminum copper interconnects were replaced with electroplated copper (Andricacos et al., 1998).

In the 1980s, a fundamental approach was taken to study the formation of epitaxial silicon on a crystalline silicon substrate. Meyerson (1990) deposited silicon by chemical vapor deposition in an ultrahigh-vacuum system at temperatures in the range of 500°C, thereby limiting impurities and significant movement of atoms in a layer and between layers. This procedure was then extended to the introduction of dopants and to the formation of silicon:germanium (Si:Ge) alloys. The Si:Ge alloys in heterojunction bipolar transistors are used in mixed-signal circuits.

(Many additional papers on this topic are included in special issues of the Journal: J. Res. & Dev. 25, No. 5, 1981; J. Res. & Dev. 44, No. 1/2, 2000; J. Res. & Dev. 44, No. 3, 2000).


Oct 03 2007, 03:58:47 PM EDT Permalink

Papers on Applications of Information Technology

In “A Framework for Information Systems Architecture” (Systems Journal, 1987), J. A. Zachman stressed the importance of information systems architecture in bringing structure to complex and decentralized information systems. Zachman introduced a framework to describe a complex IT implementation and define its architecture. The Zachman framework, in a modified form, is still being used by many IT professionals for purposes such as assessing the completeness (vis-a-vis an organization's information needs) of software development process models.

A more recent infrastructural advance is that of the Web Services standards, which facilitate the rapid development of lightweight yet powerful applications through the use of services which are hosted on remote systems and accessed on demand over a network. The tutorial paper “Introduction to Web Services Architecture” (Systems Journal, 2002) by K. Gottschalk, S. Graham, H. Kreger, and J. Snell introduces the Web Services architecture and its components and describes them in the context of service-oriented architecture (SOA). The SOA approach helps automate the management of business processes and their dynamic modification in response to changes in the business environment.

Other papers in this section describe the modeling of business processes and the implications for (and the role of) IT assets in the transformation of enterprises to on demand businesses.


Jul 31 2007, 03:53:03 PM EDT Permalink

Papers on Software

In the area of software, the 50^th anniversary report of the IBM Journals profiles several papers that describe contributions of historical significance made by IBM researchers. We highlight here two such papers: The first covers high-level programming languages; the second covers language processor technology.

In “History of IBM’s technical contributions to high-level programming languages,” J. E. Sammet describes four programming languages developed at IBM that have had a major impact on the industry: APL, FORTRAN, GPSS, and PL/I. FORTRAN is still the most widely used programming language for scientific, engineering, and mathematical problems.

In the related paper “The history of language processor technology at IBM,” F. E. Allen describes the evolution of compiler technology from the early 1950s to the late 1970s from an IBM perspective. Here, too, FORTRAN takes the center stage; it was the first language that demonstrated that efficient object code could be produced by a compiler.

The work on FORTRAN was started in 1953 by John W. Backus with the idea of building a programming system that makes the writing of code easier. Backus led an IBM team that by mid-1954 produced a draft specification for the IBM Mathematical Formula Translating System, whose name was abbreviated to FORTRAN. The first FORTRAN compiler, delivered in 1957, was an optimizing compiler in order to produce code whose performance was comparable to that of hand-coded assembler language. The use of FORTRAN reduced the number of programming statements by a factor of 20 and quickly gained acceptance.

After almost 50 years of use, a vast body of FORTRAN scientific and engineering applications has been accumulated. FORTRAN is the primary language for some of the most intensive supercomputing tasks, such as weather and climate modeling, computational fluid dynamics, and computational chemistry. Starting in 1966 standard versions of FORTRAN have been defined, the most recent one being FORTRAN 2003. Work is under way to develop a revision to FORTRAN 2003, tentatively named FORTRAN 2008.


Jun 29 2007, 03:33:05 PM EDT Permalink

Storage Systems and Databases

No doubt, you’ll be able to find other extremely important papers, and we invite you to bring these IBM Journal papers to our attention and let us know your reasons for considering them to be among the most significant papers published by the IBM Journals. We also welcome your personal reminiscences on the evolution of information technology over the past fifty years.

Papers on Storage Systems and Databases

Our section on “Storage Systems and Databases” covers a range of technologies, from the 1957 paper describing the first magnetic hard disk for data storage to the rise of relational databases. (Note that a separate “Hardware” section includes additional papers on tape devices and the evolution of magnetic storage.) The following provides an example of a significant paper that concerns hard disks and another that concerns databases.

In the very first issue of the IBM Journal of Research and Development, T. Noyes and W. E. Dickinson published “The Random-Access Memory Accounting Machine—II. The magnetic-disk, random-access memory.” This 1957 paper describes the first magnetic hard disk for data storage, thus heralding a new era the history of computing. Developed in San Jose, California, the 305 Random Access Method of Accounting and Control (RAMAC) permitted random access to any of five million bytes of data stored on both sides of 50 two-foot-diameter disks. The magnetic hard disk was adopted throughout the industry. In 1960, the RAMAC 305 was used to score the Winter Olympics in Squaw Valley, California, and to tally votes at both U.S. political conventions. Despite the high cost, the movable read–write heads of RAMAC finally made it practical to provide interactive computer systems. In 2005, the Magnetic Disk Heritage Center and IEEE honored the IBM RAMAC 305 with an IEEE Milestone Award.

In 1976, D. Chamberlin and colleagues published “SEQUEL 2: A unified approach to data definition, manipulation, and control,” in the IBM Journal of Research and Development. SQL (Structured Query Language) is the most widely used database query language. Originally known as SEQUEL, this language has been used by IBM, Oracle, Microsoft, and other major vendors of database software. In 1987, the International Organization for Standardization (ISO) adopted this language as an international standard. SQL is the unifying standard around which the multibillion-dollar relational database industry has developed, and it provides the main user interface for DB2^®, a family of relational database products offered by IBM. The 1976 IBM Journal paper, which is very highly cited, was the first publication to describe the full functionality of the language as we know it today.

Other papers are also highlighted in this section on “Storage Systems and Databases” and include such topics as evaluation techniques for storage hierarchies, the database language “Query-by-Example,” storage and access in relational data bases, an overview of IBM DB2 and methods for maximizing its performance, and the evolution of storage facilities in AIX^® for RISC System/6000^® processors.

We note that modern commercial computer applications often require rapid access to large amounts of data in the form of databases. Databases not only represent a crucial infrastructure for computer applications, but they also process the transactions and exchanges that support the economies of numerous countries. Many modern markets rely on relational databases based on models proposed in the late 1960s and early 1970s. These models permit significant data independence by providing a logical view of the database that avoids details of the physical storage. Today, most database structures are based on the IBM concept of relational databases.


May 30 2007, 04:42:19 PM EDT Permalink

Fundamental Science and Technology

Papers published in the IBM Journals are both similar to and yet different from papers published in scholarly technical Journals of the IEEE, ACM, or American Physical Society, for example. A similarity is the peer review process which is essential for all scholarly journals. After an initial editorial review, papers submitted for publication in the IBM Journals are reviewed by three or more experts in the technical field. Only papers recommended by the reviewers are published. Authors revise papers based on the reviewers’ comments and these papers are then edited by both technical editors and copy editors in order to produce high quality technical publications which are understandable to as broad a readership as possible. The IBM Journals report advances in information technology from fundamental science to hardware and software technology to systems, services and applications. The large range of subjects covered in the IBM Journals distinguishes them from many technical society Journals.

The breadth of the IBM Journals is indicated in the 50^th anniversary report where papers are listed in the following categories: applications of information technology, storage systems and databases, computing system architectures, computing methodologies, software, hardware design and implementation, device materials and processes, and fundamental science and mathematics. In this note we would like to comment on some papers in the section on fundamental science and technology.

Papers on Fundamental Science and Technology

Fundamental advances in basic science tend to be initially published in the Journals of particular scholarly disciplines or in recent years in Science or Nature. Nevertheless, several examples of initial publication of significant science in the IBM Journals are given in the special report celebrating 50 years of the IBM Journals. Here is one example in physics and one in computer science.

In the third issue of the IBM Journal of Research and Development in 1957, Rolf Landauer published a paper on “Spatial variation of currents and fields due to localized scatterers in metallic conduction”. In this paper Landauer provides a solution to the transport equation for electrons in metals in the presence of localized scatterers. The significance of this result was not immediately recognized, but subsequently it has been cited hundreds of times. Although the Landauer conductance was not actually measured until the 1990's, it is now part of all discussion of very small devices—particularly conduction in molecules, carbon nanotubes, semiconductor quantum dots, and the very narrow semiconductor nanowires that people are trying now to build. The Landauer formula is now known and to be the simplest and most natural method for calculating conductance in such small structures.

In 1959 Arthur Samuel published a computer science paper in the IBM Journal of Research and Development on “Some studies of machine learning using the game of checkers”. He showed that an IBM 704 computer could be programmed “so that it will learn to play a better game of checkers than can be played by the person who wrote the program”. This was the first program of its kind and was a milestone in artificial intelligence programming. The machine learning principles developed in this work were steps toward the ultimate match between chess champion Gary Kasparov and the IBM Deep Blue computer which showed that machines could learn to play complex games and were capable of defeating even the most formidable human opponents. Today, machine learning has countless applications in fields that range from search engines and medical diagnosis to bioinformatics, stock market analysis, and robot locomotion.

The IBM Journals also feature papers which review significant fundamental work. For example, in 1986, G. Binnig and H. Rohrer published “Scanning tunneling microscopy” in the IBM Journal of Research and Development. This paper reviewed the physical basis of this technique which provides images of the surface structure of materials with atomic resolution, and the instrumentation required for a scanning tunneling microscope. It included studies “which suggest potential applicability of this new type of microscope to a relatively broad range of biological, chemical, and technological areas”. For their invention of scanning tunneling microscopy, G. Binnig and H. Rohrer received the Nobel Prize in physics in 1986.

The special report celebrating 50 years of IBM Journals includes papers on high temperature superconductivity, laser science, computations in atoms and molecules, cryptography, the thermodynamics of information processing, algorithmic information theory, and other topics. We invite your comments on these and other papers published in the Journals.

In future notes we will highlight papers published in the Journals in many of the other important areas of information technology.


Apr 19 2007, 05:48:44 PM EDT Permalink