Abstract:
Financial industry and manufacturing are the main customers in migrating existing
applications into client server environments. This paper helps in identifying the candidates
for migration and suggests a method for migration. Business evolution and existing
application architecture influence its sequence of steps. The most frequently applications
moved down from mainframe solutions are: sales, order entry, material requirements
planning, executive information systems, finance, ledgers. Re-engineering will proceed in
the following sequence in a one-by-one move from an old application to a new one, but
never in an all-at-once go.
1 . The application logic is split into objects, wrapped and re-allocated according to
businesses processes across several platforms
2 . The data locations and movements follow the application logic-not as previously, where
a quasi-'eternal' enterprise database dictated business processes and application logic.
Those are the migration steps:
1 . Quantify the added value of each migrated or newly computerized function, its
performance gain and resulting cost reductions
2. Understand and document the business reasoning behind each application. Deduce the
high level application logic from the business reasons and logic. Start with the external
business context.
3. Differentiate the functions of the application logic in order to prepare for a split. (e.g. in
an airline crew scheduling the separation between local data access, data storage and
update suggested an effective split)
4. Does the current application reflect the business logic? If not, redesign from scratch.
5. Decide on the split between server and client according to the guideline below.
6. Exploit the modularity of existing applications in order to attribute front-end interface
functions to clients and back-end database functions to servers.
The guideline for choosing the right server will consider the work situation and the number
of users per servers: For personal use up to 30 users the server will be on a high-power PC
or a R I SC workstation. High power RISC machines will l serve work groups and large
department up to 100 users, while Superminis and Database machines support large
enterprises up to 1000 users per server. The cheapest expansion of server performance is
most often a parallel processor.
Sizing of servers and client machines will follow two iterations: firstly, balancing system
component capability, load parameters and performance goal and secondly, optimizing
design.
From the beginning of any re-engineering effort it must be borne in mind that small LAN
servers have their weak point in communications during development and operation while
large distribution projects have their bottleneck on operational management. Standards for
management of applications, transactions, system and data administration are absent or
rudimentary only. Standards for network management, however, exist on the level of
network management protocol(SMN P for TCP/IP), ISO API for management tools, common management interface protocol(CMIP) and vendor consensus(H P and I BM) for large systems. The management support services should include minimally: echoing test
messages, transaction status reports, log of activities and automatically documenting the
updating software releases, tracing facilities.
