MIGRATING ADMINISTRATIVE APPLICATIONS TO A UNIX/RDMS 
ENVIRONMENT

CAUSE94 Current Issues Session
Thursday, December 1, 1994, 4:30 to 6:00 pm
Walt Disney World Dolphin, Oceanic Room 6


Moderators: Ken Maharaj, Dalhousie University and 
            Robert Velasco, University of Arizona 

CURRENT ISSUES SESSION ABSTRACT

Are you "downsizing" to Unix? Which RDMS did you select? How 
will you set about migrating one or all your administrative 
data and operations? Will you be using a GUI for 
your applications? Which one? What application tools will you 
use? How will you deploy and maintain your applications? Will 
you need to acquire a transaction processor? Will you retain 
your mainframe as well, and pass data from one to the other? 
Will your databases be located on one or more than one 
machine? What kind of people and training will you need to 
accomplish this?  Lots of questions! If you have or are 
looking for some of the answers, and are willing share them 
or hear them, be sure to drop into our session. A special 
welcome if you are a subscriber to the MVS-to-Unix listserv!

CURRENT ISSUES SESSION SUMMARY

List of Attendees:
Twenty-two people recorded their attendance:

Gaston Blais		University of Montreal
Michel Vanier		University of Montreal
Danny Kemp		Miss. State University
Natalie Vincent		Syracuse University
Barbara Germon		Emory University
Sandy Dennhardt		NorthWestern University
Terry Cowan		University of Wisc. (Milwaulkee)
Marge Waala		University of Wisc. (Milwaulkee)
Ron Dawe		Colorado State
Norm Hadley		Middlebury College
David Lassner		University of Hawaii
Barry Walsh		Indiana University
John Rogers		Dalhousie University
Rob Alworth		L.A. Community College
Htay Hla                University of Arizona
George Ellis		University of S. Florida
Scott Higgins		Baylor University
Allen Lind		Board of Governors University
Mickey Green		Princeton
Michael Rossi		Polytechnic University
Mahbuba Ferdousi        Emory University
Ken Maharaj             Dalhousie University

We quickly decided to make an agenda for ourselves, consisting 
of issues about which we to relate our experiences - or raise 
questions. Although most of the group were not on the 
MVS-to-unix list, and had not seen the "list of migration 
issues" published on that list, the issues identified corresponded 
quite closely the latter.

The issue of people training and change of culture came up 
early. One member related that the unix personnel at their 
site did not understand the need to keep the "system" up and 
online at all times. They tended to take the system down 
whenever it suited them, without notice. The switch to 
"distributed thinking", and Windows-based programming was 
also contrasted with centrally-based programming and 
operations. It was agreed that significant funding and time 
needed to be allocated for personnel re-training.

We discussed the "rationale" for making the switch. Some 
attendees said that their institutions planned to replace 
their mainframe entirely, with attendant savings in 
operating costs. This involved a complete replacement of 
the the application portfolio, a difficult and expensive 
undertaking, but it was felt that there were separate 
paybacks associated with doing that. Other attendees noted 
that they do not plan to replace their mainframes with unix 
machines, but intended only to take advantage of the 
capability that these afforded for incremental capacity 
growth. Significantly, everyone agreed that cost 
justification was only one type of rationale: the other was 
the usefulness of the new environment to the end-user and 
to the enterprise, generally.

Because some sites intended to carry out unix operations 
in parallel with traditional mainframe ones, they felt 
that techniques for data migration and dynamic data transfer 
between the two types of operation, were very important. 
We did not get an opportunity to explore the actual technical 
approaches being taken to achieve this.  The topic of 
Application Development Tools and Training was put on the 
agenda, but we did not get to discussing this one in detail. 
We did expand the topic by noting that application design, 
deployment and maintenance were all different under a 
distributed environment, and required major re-thinking.

Standardizing on a relational database was seen as one of 
the essential parts of the overall architecture. The relative 
merits of the leading databases were discussed briefly; one 
attendee wrapped up the discussion with the somewhat 
surprising and reassuring observation that they had just 
migrated a database from Oracle to Sybase, and that it was 
"basically a non-event". Another point of differentiation 
for various sites was the extent to which they chose to 
adopt the RDMS as a "live" transaction-handling tool, versus 
an off- line repository for DSS. There was a strong feeling 
that these uses should be separated, with different databases 
being defined.

Control of operations was another key point of interest. 
It was noted that in the unix world,  many operational 
subsystems which were taken for granted in the mainframe 
environment (eg., file handling, disk storage management,
job queues, print control, etc) had to be separately found 
and installed. Coupled with this was the physical security 
issue. It was felt by some that CA's Unicenter offered the 
most complete solution for these issues; others were 
accumulating separate solutions. Although security was an 
issue that often seemed to present particular difficulty 
in migration to unix, more than one university's legal 
department had pointed out to attendees that the basic 
requirement was "due diligence" in the protection of 
data and access.

Performance was brought up as a vital issue, which I 
do not think was recognized on our original mvs-to-unix 
list. It was pointed out that analysis of performance 
problems was going to require its own set of skills and 
tools: in the new environment, there was the potential 
for performance problems at a number of points, namely,
at the workstation, on the network, on the servers, or 
at the database...

Outsourcing was raised briefly as an option that had 
been used, both during a transition, and afterward.

Finally, another sleeper issue in this transition is 
that of data policy, and custodial policy. At Dal, we 
had recognized the data problem early on, and moved to 
establish a committee which would sponsor data as a 
corporate, not a private, resource, available to any 
employee legitimately carrying out their work; but we 
hadn't explicitly anticipated that the daily processes 
themselves, as they became segmented and redefined, would 
be a contentious piece of fiefdom. Our department has 
decided that we must create processes which can be 
executed wherever convenient within the organization - 
without ourselves necessarily getting involved in 
struggles over this. Further, the ideal situation would 
be wherever possible to try to build processes that do 
NOT actually require client custodianship, (ie, just 
like EMail, or word processing).

Several attendees had real projects in the works. Htay 
Hla said that they were moving applications from a 
Prime to unix: they were looking for a stable long-term 
environment for the long-term; another attendee reported 
that their strategy was to create a Sybase warehouse for 
a data dictionary and for "reusable business objects"; a 
third (Princeton?) was moving their HR system fro the 
mainframe to a Sybase server; Dalhousie planned to migrate 
all of its administrative applications to Ingres on unix; 
there were various projects under way at Indiana 
University; and the University of Montreal was migrating 
from an obsolescent Cyber to unix.

All in all, we had more than an hour of cordial and 
intensive exchange on a topic we were all involved and 
interested in! Several people commented that it would be 
nice to have this meeting every year.

Ken Maharaj
Director, Adminitrative Computin g Services
Dalhousie University

Co-chair of Special Interest Session
(Htay stood in for Robert Velasco, who had a scheduling conflict)

Date Revised: 2/15/95 (rlr)