Readers Respond: Campuses that Eliminated Mainframes or Plan To and What Hardware Configuration Replaces Them *************************************** Copyright CAUSE. Permission to copy or disseminate all or part of this material is granted provided that the copies are not made or distributed for commercial advantage, the CAUSE copyright and its date appear,and notice is given that copying is by permission of CAUSE, the association for managing and using information technology in higher education. To disseminate otherwise, or to republish, requires written permission.For further information, contact CAUSE, 4840 Pearl East Circle, Suite 302E, Boulder, CO 80301, 303-449-4430, e-mail info@CAUSE.colorado.edu *************************************** Readers Respond Question: Has your campus recently eliminated any mainframes, or is there a plan to do so? If so, what kind of hardware configuration has replaced/will replace them? Bryn Mawr has replaced its super minicomputers with Unix systems. The former VAX/VMS timesharing environment has been replaced by DEC/Ultrix and Sun SPARC servers--with separate machines for communications, stat packages, FORTRAN-based number crunching, visualization-graphics, and Mathematica kernel. E-mail and TCP/IP services are run from client software on desktop systems. Small-scale calculations have also migrated to the desktop. The administrative applications environment (AIMS, from AGI Software, Inc.) has been ported from a Primos/Prime Information environment to run now on HP-UX/Vmark Universe. The new host computer is an HP 9000-847. Most new administrative applications are based on office LANs. Thomas A. Warger Director of Computing Services Bryn Mawr College (PA) twarger@cc.brynmawr.edu Blackburn College, a four-year liberal arts institution of about 500 students, has migrated computer operations from minicomputers to local area networks. Previously, two minicomputer systems--a DEC VAX 11/750 purchased in 1983, and a used Hewlett-Packard 3000 Series 42 purchased in 1990--had been used to support both academic and administrative needs. A new computing strategy was adopted last summer, calling for two separate PC-based local area networks, one for administrative needs and one for academic use, both running under Novell NetWare 3.11. The first phase of the plan saw the implementation of the administrative network, installed by the hardware vendor. A 33 megahertz, 80486 PC-compatible with 12 megabytes of memory and a 500 megabyte hard disk is the file server for a 10BaseT Ethernet network of 20 workstations, mostly 386SX computers running an integrated administrative software package from a third-party vendor. The second phase of the plan coincided with a complete renovation of Blackburn's academic computing center. The VAX and HP3000 were shut down last December to allow demolition of the old computing center facilities. The old computing lab was converted into three separate labs. After the equipment was moved in and set up in the new facilities, the third phase was begun: another PC-based local area network was implemented to connect all the PCs in the labs, plus those of several staff and faculty. Computing center personnel have seen an upside and a downside to the downsizing of their computing platforms. Some of the benefits have been the elimination of costly air conditioning requirements and maintenance contracts with DEC and Hewlett-Packard. On the other hand, the MS-DOS and Novell NetWare operating systems leave some things to be desired by those used to managing mainframe operating systems. Printing has not yet proven to be consistent and dependable on either network. Because batch jobs scheduled for off-peak hours must now run on DOS workstations, they lack the security to which we have been accustomed. Users of the administrative system have also experienced a frustrating series of problems in migrating to the new software. We have read similar stories of small to medium sized companies and institutions replacing their mainframes with PC LANs, and they seem to share the common theme that it takes longer, costs more, and is far more difficult than the initial projection. We have learned that the complexity of tying together several generations of hardware and software is more difficult to manage than we expected. For example, we are being driven toward upgrading and replacing both hardware and software that we thought we would still be able to use. Even small upgrade costs become large as they are multiplied by the number of units involved. Jack Anderson Director, Computer Center Blackburn College (IL) 217-854-3231 Dalhousie University is migrating its central administrative applications from the existing MVS IBM 4381 platform to a client- server environment with the servers being Unix systems. This is a three-year project which is due to be completed in 1996. The first Unix system purchased is an IBM RS/6000-580 computer running AIX. The Ingres RDBMS and several Ingres tools are being used in the target environment. Ken Maharaj Director of Administrative Computing Services Dalhousie University (Nova Scotia) kmaharaj@adm.dal.ca At Carnegie Mellon University we started the elimination of our "mainframes" in early 1988. It is a little hard to tell what a mainframe is these days in any case. What we eliminated over a three- year period were three physically large, technologically outdated, severely underpowered, and ridiculously expensive information processing systems. These were: an IBM 3083, two DEC 2060s, and at the end of the process a DEC VAX 8700 and a VAX 6440 for a grand total of 40 CISC MIPS. The price range on these configurations was $2.5 million to $500,000 including so-called higher education discounting. The CMU cost was considerably less since the IBM 3083 was a "grant" machine and the cost of the VAX6440 was 60 percent underwritten by the sale of the VAX 8700. These were mainframes in the real sense of life-cycle cost: expensive to buy, expensive to maintain, and expensive to support, and with a residual value profile that became, and is today, impossible to justify to any CFO. The four current CMU configurations combine for 14 times (575 MIPS) the CISC MIPS output of the previous "mainframe" systems complement, and drive 20 times (1 GByte) the main memory and 15 times (75 GBytes) the disk storage of the previous systems. The net total purchase price of these four configurations was less than $500,000, the maintenance cost of the hardware is less than 25 percent, the internal systems support less than 25 percent, and the residual value ratio (purchase cost to market value after one, two, and three years ) is more than 1.5 times any of the previous configurations .The four machines are: DG/AVIION 8000, Sequent 2000/250, Sequent S-81, and DEC VAX 6610. All machines are upgradable in their current cabinetry so as to achieve 4X to 10X incremental enhancements in capabilities at less than 50 percent of the original purchase price. In sum, CMU is running a greatly enhanced processing environment at extensive savings, all of which was achieved by early recognition of the decline and demise of the mainframe and by the implementation of a transition plan which took advantage of funds released from the elimination of the overly expensive computing hardware configurations of the time. Len Brush Assistant Vice President Administrative Computing / Information Services Carnegie Mellon University (PA) lb1z+@andrew.cmu.edu South Dakota School of Mines & Technology, located in Rapid City, South Dakota, offers BS, MS, and PhD programs in all major science and engineering fields. Current enrollment is approximately 2,300 students. Our campus has utilized various CDC mainframes to serve academic, research, and administrative computing since 1969. In the last seven years, we have seen a slow migration by research and academics to workstations and PCs. We currently have a CDC Cyber 180/830 running administrative accounting and some academic instruction. We will be moving the administrative computing to a centralized Regents Information System on an IBM ES9000 within the next year. Simultaneously, we will move the remainder of the academic computing to an RS/6000. It is our observation that the network is becoming the heart of our computing environment. Distributing the computing power has worked only because we have been able to continue to provide access to it via the network. Jerl E. Pringle Director, Computing and Networking Services South Dakota School of Mines & Technology jpringle@msmailgw.sdsmt.edu The Citadel doesn't have an IBM-type mainframe, but we do have a two- node, HSC-based VAX cluster that was purchased in the late 80s for about $1.5 million. This cluster includes a VAX 6420 and an 8530, and we've been very happy with the performance of this system and the variety of VMS software applications we've been able to buy for it. But so many people are using the VAXes now that they're beginning to slow down during periods of heavy CPU usage. How can we solve this response time problem, given that we have very little money? Our initial plan was to buy a low-end Alpha machine (DEC 4000) and integrate it into our existing VAX cluster. But this "mixed cluster" approach turned out to be too complicated. So ... we are simply going to "downsize" our entire VAX cluster by replacing our two VAXs, HSC70, and 12 RA disk drives with two VAX 4100s, a DSSI bus (cabling), two DSSI-to-SCSI adapters, and eight 1.6 GB SCSI drives. The total cost of this new system will be about $100,000, and our VAX- related hardware and software maintenance payments will drop $50,000 a year: a two-year payback. But the CPU power of the new DSSI VAX cluster will be almost three times greater than our current cluster, and we can add a third VAX 4000 for less than $50,000 if we need more power down the road. Now all this may seem like small potatoes to colleges that have true mainframes, but I think it's important for those who have much smaller systems to seriously consider even more economical options. Downsizing (more accurately, downcosting) is not just for mainframes. Rod Welch Director, Information Resources Management The Citadel (SC) welchr@citadel.edu What if we were to look at the downsizing phenomenon from the viewpoint that mainframes aren't being eliminated, they are just becoming the biggest servers on our networks? This strategy allows us to keep legacy systems functioning while we proceed to future platforms. If the mainframe were called the "mainserver" or "principal node," would there be so much tumult? The University of Miami's clustered VAXes act as a mainframe, although bought as separate units. Is the real problem that the big processor is under one box, one control, or one person? In my view, the position I call the CINO (chief information network officer) actually has more technical power than the CIO. Lew Temares Vice President and CIO for Information Resources University of Miami (FL) m.lewis.temares@umiami.ir.miami.edu An internal review of administrative computing and information technology architectures at the University of Manitoba noted that computer services budgets have been declining in terms of purchasing power; demand is increasing for administrative computer applications development; support, maintenance, and development of existing systems continue to be expensive in terms of staff time requirements; and there have been fundamental changes in computing and networking technologies and economics in the past two years which make available new alternatives in administrative computing. The results of this internal review indicate that the time is right for the University to take advantage of new alternatives to maximize the return on computer services staff time and to further empower our clients by providing them better tools to access and analyze pre-defined collections of their data. The report recommends a review of DBMS and mainframe computing platform alternatives resulting in acquisition of alternative hardware and software; development of a "data warehouse" of extracted corporate data; definitions of tools and standards; a study of application development methodologies; and defining a time frame for converting existing systems to alternative open technologies. This report has been accepted by the administration and we are embarking on some of the activities. The first, a review of DBMS technology, has been completed and we have chosen Sybase. There is an ongoing review of hardware and other software to be completed by the fall. The next step will be designing the data warehouse. Actual migration of code has a lower priority. Our goal is to be off the Unisys mainframe, which now runs a library automation system, by 1995 and off MVS by 1998. Finally, it's important to note that the primary motivation for migration is both financial and the desire to improve the user interface and the productivity of application developers. Gerry Miller Director, Computer Services University of Manitoba (Canada) gerry_miller@umanitoba.ca With strong support from our president, Lock Haven University (3,900 students) has come tantalizingly close to the planned goal of retiring our 4381 mainframe. As part of our plan we have abandoned all "legacy" IDMS/R-based systems, opting instead for new integrated packages in a client-server enterprise architecture. New applications implemented include the admissions, financial aid, housing, student billing, student records, advising, and alumni/development modules from CARS Information Systems Corporation. Simultaneously we have implemented the Dynix library automation system. All these applications were implemented on a fiber-optic campus backbone network which connects each of our 25 buildings to the network control center and our four IBM RS/6000 network hosts. Using three Novell servers and TCP/IP client software, our clientele enjoy instant access to programs and information using a "point and click" Windows interface. We have begun to recoup the planned savings by not renewing our IDMS/R maintenance this year. The remaining challenge is to implement a successor to our financial application. The last mainframe application, this will close the book on a chapter of computing history at Lock Haven. Robert O. Little Director, Computing Center Lock Haven University of Pennsylvania rlittle@eagle.lhup.edu The Minnesota State Universities at Bemidji, Metro, Mankato, Moorhead, St. Cloud, Southwest, and Winona will replace a central Unisys 2200/611 administrative computer with a client-server system. Digital VAXes will be located at each university connected to LANs and to MSUSNet, the system's data network. Unisys computers have been used for years with each university responsible for developing its administrative applications. The new approach will distribute VAX computers to the campuses but will use a shared approach for the database development. The common student records system will be reengineered with fiscal and human resources systems converted. Equipment will be installed this July. Don Olson Associate Vice President of Computer Services Mankato State University (MN) don_olson@ms1.mankato.msus.edu For almost twenty years, Western Oregon State College's administrative computer needs were met through a remote job-entry terminal to a central mainframe computer that served all four regional colleges in Oregon. It has been only during the past three years that the mainframe hardware and software necessary to serve the student information needs of the campus were acquired and put into service. Western has utilized microcomputers almost exclusively for instruction, with the exception of two small mainframe Sequent computers which were donated to our instructional program two years ago. Bill Cowart, Provost Western Oregon State College cowartb@fsa.wosc.osshe.edu In November of 1992, Arizona State University's four-year lease on a CRAYXMP/18se running UNICOS expired. Budget reductions during the three preceding years encouraged us to reconsider how we were going to support research computing in the future. The lease plus maintenance amounted to approximately $1 million a year, and only the maintenance stream (about $250,000) remained going into the 92-93 fiscal year. A "transition team" was formed to look at jobs on the CRAY. It was determined that most of the codes were double precision, floating point, requiring large memory. The University advisory committees had already determined that the future direction for computing oncampus was to be distributed, with as much of the resources as economically feasible moved as close to the desktop as possible. Consequently, using actual CRAY jobs, the transition team began the task of benchmarking platforms from the five major workstation vendors. A decision was made to purchase five IBM RISC/6000 systems. Three were model 580s with 256mb of memory: one for interactive use and the other two for batch activity. The fourth was a model 530, which was added as a file server, and the fifth was a model 520 for use in development and testing. The systems were installed in early November, giving us about one month for the transition. This phase went off without a hitch and the CRAY was removed by the end of November. Our experience has been that while many of the jobs are slower on the IBM workstation, the fact that we have three such stations has resulted in increased throughput. Coupled with the fact that we can incrementally add systems to increase our effective resources, this has been a "win" scenario for ASU. William E. Lewis Vice Provost for Information Technology Arizona State University william.lewis@asu.edu Bradley University is midway through an initiative to replace our mainframe computer with microcomputer networks. We have identified three families of applications: advancement, financial, and student records. Each of these families will use one or more network servers. The advancement application will be converted to a client-server architecture in June 1993. We developed this software in-house. The software is based on PC-compatible computers, but will accommodate Macintosh computers and will support dial-in access. For the financial system, we purchased a client-server fund accounting system from a software vendor. We will be installing this system over the next fiscal year. Personnel/payroll will be live January 1994 with the rest of the system at the start of FY94-95. We will begin working on a client-server-based student records system once we complete the advancement project this year. We expect to be working on the student system until mid 1995. Stephen Patrick Director of Computing Services Bradley University (IL) steve@bradley.edu The 1990 Lehigh University Computing and Communications five-year plan included the introduction of workstations at an accelerated rate, extending the campus network, elimination of mainframe computers, installation of compute and file servers, and introduction of software which would make this combination operate as a unified computer system. The benefits of this new computing strategy were much more computing power, improved computer/user interface, more advanced software, greater access to information and computing off-campus, and ability to finance with existing funds. The first step was the elimination of the old CWIS system and the introduction of an RS/6000 cluster, with new software, to take its place. Roughly, the new system cost about 20 percent of what the old system cost six years ago and the new system has more than twice the capacity. During the past twelve months we have also replaced two academic mainframes (CDC Cyber and VAX) with powerful IBM RISC computers. This replacement has resulted in a distributed, client-server system, with over ten times the computing power at one tenth the cost of the former mainframe systems. Lehigh is now extending the distributed environment by installing approximately 150 IBM RS/6000 workstations, sharing file systems, using the Andrew File System. The computing power available to the campus has increased by about 500 times, with no increase in budget, and the functionality of the systems has been greatly increased by the superior software now available on these systems. William Harris Director, Computing Center Lehigh University (PA) wrh0@lehigh.edu Who wants to get rid of mainframes? Getting rid of mainframes could be a godsend to budget-limited organizations no longer able to keep up with feeding the beasts. At New Mexico State, the interest is zero outside of the computing organization but of significant interest inside. In order to interest the overall organization in moving in this direction, two major client-server applications are now in progress. The first, a workstation-based data warehouse, would provide access to three to four now incompatible data systems currently on the same mainframe (a kind of enticement). Running the huge, new student system on a similar workstation, now nearing completion after five years of construction, is approved as a move toward rightsizing. Note which organization is doing the pushing--the computer center. Success in policy is, therefore, not guaranteed. What enticements would be more promising? Mainframes are wonderful for those who do not have to pay for them or even worry about those who do pay. Policy now is that all new applications, locally constructed or purchased, by default, go to the mainframes since that is a "free service." No billing is done for administrative applications, even when they overrun allocated costs. Under this approach, there is absolutely nothing wrong with mainframes. At least two different enticements could reverse this attitude. One is putting the clients on the offensive to get rid of mainframes. The easiest way would be to split up this central budget for mainframe support among these clients and let them find the optimal cost- effectiveness in running their applications. A second would be to use chargeback approaches in a realistic manner and limit services to the resources available. Both would produce cost-consciousness, perhaps the only viable method for downsizing. What is the prognosis for client-server applications? Getting rid of mainframes is not a priority for clients unless they have to foot the bill for their service. Since they don't have to pay or account for costs, the mainframe is a free good and no disservice to clients at NMSU. Why would they want to get rid of something from which all of their services come at no cost? Let them separate with their portion of the pot of gold, and these clients will find a better (and more inexpensive) approach. A first sign will be how well the testing and evaluation go for the two client-server applications--the parallel student system project and the data warehouse. My prognosis is they will go poorly because no investment in the process need be made by clients to save money. If it was their money on the line ... ? A second sign--will access to information be so important as to make the new directions valuable? If the clients had "called the shots," I would be more optimistic. Why do these projects then? The only answer--they are correct in concept! The outcome--questionable. Prognosis for the mainframes: they will still be here for more time than their value should allow. Joseph R. Denk Director,Computing and Networking Technology New Mexico State University jdenk@nmsu.edu Syracuse University is committed to a phased migration to client- server computing over the next five years. The university's VAX 8820, one of our primary academic time-sharing systems supporting 4,500 accounts, was removed from service in May, culminating a year-long migration effort. A small number of VAX users are moving their work to desktop systems, but due to the lack of acceptable remote-access facilities for desktop-oriented electronic mail systems, the majority are migrating to our distributed client-server Unix environment, called SUnix. SUnix consists of workstations, timesharing systems, compute servers, and other specialized servers (e.g., mail and news), all connected to a central server that houses user directories. Advantages of a Unix-oriented client-server environment, compared to that of traditional mainframes, include a more diverse suite of available applications, superior compute cycle price performance, enhanced integration with desktop computers, and scalability for supporting an increasing user base. This spring, we announced plans to decommission in May 1995 our IBM 3090, a VM-based academic timesharing system with 15,000 user accounts. Our overall analysis of usage patterns suggests that our greatest challenge will be to find alternative platform solutions for two broad classes of users: those who depend upon the 3090 for electronic mail and those who use the system for statistical analysis of large datasets. We are optimistic that within the next two years, functional and scalable remote-access solutions that allow e-mail users to migrate to desktop platforms will emerge. For users with large-scale statistical analysis requirements, we are confident that the combination of more powerful desktop computers/operating systems and RISC-based compute servers, together with near-line storage systems, will provide computational, I/O, and storage capabilities that offer functionality superior to that offered in the existing mainframe environment. We are also preparing to remove our administrative mainframe, an IBM ES9000, probably in 1997 or 1998. We are just beginning to develop pilot applications using Unix-based database server technology in combination with network-attached IBM-compatible and Macintosh microcomputers. Ben Ware Vice President, Research and Computing Syracuse University (NY) brware@syr.edu At Iowa State University we like to think mostly about good business and information systems services criteria for supporting Central Systems (referred to by many as mainframes). The elimination of so- called mainframes (which implies something negative about Central Systems) is not a major issue at ISU. Instead, we think more along the line of how to evolve the solid principles (standards, integrity, accuracy, fast response time, up-time, cost containment, planning, access, security, etc.) of strong Central Systems to something better. We refer to our so-called mainframe as our Central Server and indeed the system (more than one box ) is the main data and network server. One of our main goals is to provide solid customer-oriented information system services that utilize the many good aspects of central systems, networked systems, LANs, client-server approaches, workstation orientation, and user friendliness within the funding and priorities of Iowa State University. With current technology, we are using many servers that are networked at various levels to best meet our goals. The Central System is still the strong focal point for the information systems. We envision this to continue, while the architecture of the Central System certainly can and will change depending on technological and economic developments. Wayne Ostendorf Director, Administrative Data Processing Center Iowa State University woosten@isuadp3.adp.iastate.edu ====================================================================== Two electronic discussion groups have been formed by CAUSE members on topics related to this question. If you would like to join Ken Maharaj's discussion on migrating to Unix from MVS, send e-mail to MVS-TO-UNIX-REQUEST@AC.DAL.CA including your name, institution, and e-mail address. If you are interested in discussing academic computing downsizing, send e-mail to LISTSERV@ASUACAD including the one-line message: SUBSCRIBE DOWNSIZING YOUR NAME ======================================================================