CS 446 Notes 2, 30Sept 1996

Experimental Seminar on Large Scale Software Construction.

Notes for CS 446-2 Fall Quarter 1996-1996 Gio Wiederhold and David Luckham.

Admin notes: We will use the web and email a lot for communication.
My email is: Gio@cs. Let me know in the header that's about CS446
The secretary is Marianne Siroker@cs.
Project choices discuss on Wednesday Oct. 2.

Also see David Luckham's assignment, due Oct.9th.

Administrative Computing Changes at Stanfors

Mary McMillan
See handout notes,
Came from MIT adminstrative computing. There the choice was to use SAP.
SAP, a german-based company has a rather inflexibnle approach, that can however reduce (or rather amortize over more customers) mainrenance costs. .. Saw Stanford SPIRES 1979.
Mary manages now the development of Stanford University Software. Adminstrative SW is now mainframe and SPIRES-based. Over 100 shadow (or called now `local workplace systems' duplicate central support in part, those came about because inadequacies and slowness of University system. For instance. I (gio) maintain my own spreadsheets for research projects, with past expenses and future projections, because the university reports only tell me what I did in the past, and that several months late.
Central systms also include a server farm and a data warehouse. There is also a campus-wide-net with a campus-wide ID for logging in.
For new software there is an attempt to use commercial Software, hoping to reduce costs of processing. For instance, cost of an invoice at Penny's stores is $0.27 , at MIT $10.-, at Stanford?
DCL: Can Stanford influence a vendor's product? MM: yes, smaller providers, for instance the spinoff company of AMS for student finacial aid system.
Sample Systems (see handout, can we get it on line? )
  1. Office of Development.System comes from a small company. Local extensions. System size. Has ledger, keeps track of alumni. About 400 transaction type. Now in pilot use. Stanford has 250 people collecting gifts, 30 are now using the system. Enhancements can by
    • vendor contract
    • locally developed
    • collaborative with Stanford.
  2. . . .
  3. Capital assest mangenet systems combines ORACLE and SPIRES stuff to run by November. Phase 1. Programmers on project are 6 Stanford + >6 contractor.
  4. Core financials. Was to be purchased from Peoplesoft last year, now Oracle appl.suite. Does not use SQL, diect routine calls, not open. Does not follow Stanford Architeture group recommnations. Expect 53 people total, 1* Oracle, rest Stanford business types and programmers. Note in CS545 we have some ORACLE talks. Perhaps import components for workflow managment. Also integrate with SYBASE-based data warehouse. Start with a conference-room pilot. Understand best mapping to dominant business processes. Then Gap analysis between.
ITSS report available on line, lists projects.
Goups at Stanford comprise:
Assembly and Integration [McMillan]: 35 people +contracctors
Support (SPIRES) []: 45 people
Operations []: x people
Distribution and Communication []: xpeople
Architecture planning and Standards[], recommended: "move to standard Relational, eventually to object " (ignored by financials)
Customer assistance


University now has a large, complex system covering purchasing, registration email, ..., and everything. Started in 1970 with SPIRES (Stanford Physics Information Systems -- for distribution of physics preprints. Thus systems was expanded and expanded over 20 years and now is wonderful, but unmaintainable in terms of adding new functions. About 500 employees work in this organization, at $100,000+ per person it takes a lot of money.

There are certain cores: General Ledger for finance, a student recod, a library, financial investment. The system supports `Workflow' concepts: a DAG (with some cycles) describes process that a transaction goes through. Includes digital signature for validation. System is distributed, and access can come from many places. via an entrypoint: ForsytheTN. Supports RAS (reliability, accessability, security) Ras number includes all accessible time of advertised schedule: RAS number, Forsythe has 99.2% but has 30% scheduled downtime, Leland has 99.7% without scheduled downtime, uses backups to be deal with any single hardware failure. For insatnce 18 CPUs, and with 300 Gigabyte disk farm, where errors are fixed. Leland is actually 9 machines. There are 800 mountpoints, where directories are mounted when you log on. Uses Andrew file system (AFS), authentication. Update expected to NFS. Directories and files are then cached locally. Increasing use from in-house people. Leland has 19 people.

Sweet Hall has many computers for Stanford student computing. Operates in a distributed environment.

Stanford Administrative Computing to move to DCE open systems infrastructure, now being tested. None of the current developments use/are committed to DCE yet.