Milestone 7
IMPLEMENTATION & POST-IMPLEMENTATION
General Implementation Narrative:
The National Bank Exchange Bureau employs client/server architecture. At the bank level, it uses centralized architecture and at the back office level, the National Bank adheres to the client/server configuration. Inasmuch as the Exchange Bureau features commercial servers at each branch, it abides by this architecture.When fully implemented, the network topology of the National Bank features a standard Ethernet Bus topology at the local area network (LAN) stratum with a token ring topology (arm of an ADSL) internally. Furthermore, at the metropolitan area network (MAN) level, the National Bank features a high speed FDDI token ring in order to handle future capacity needs. The entire network infrastructure is quite scalable in that it is not difficult to add more branches, should this be necessary in the future. Owing to this overcapacity, this is a good opportunity to explore the Internet and/or an organizational Intranet.
Part of network implementation will comprise of the periodic migration to an Intranet implementation strategy (i.e., a move towards a more distributed architecture). When one moves towards the Internet protocol (IP), this opens up the doors to become many times more efficient, especially in terms of customer service.
The network manager of the National Bank expressed concern about a LAN that is running out of bandwidth and wants to migrate or move users, servers, and peripherals to a higher-speed network. Unfortunately, the number of competing high-speed technologies has created a lot of confusion in the marketplace. Some of these technologies are complementary, while others clearly compete with each other.
General Cost of Connectivity:
The cost of a new LAN technology should be measured in terms of connection cost, which includes the cost of the NIC and the hub port. Factor into this calculation the cost of network management software and any other new equipment that needs to be purchased.
Performance:
Performance is probably the number one concern when considering a network conversion or augmentation plan. Different vendors measure performance in various ways; here is a criterion on how to measure performance:
Wire speed and average throughput rate: There are different ways to measure performance in shared media networks such as Ethernet. The wire speed is the maximum data transmission rate for the network. For Ethernet, it is 10 Mbps, and equals the speed of the data transmission once the node gets onto the network. But since most networks have more than one user, the average throughput is significantly less for each user. One way to calculate average throughput is to take the wire speed number and dividing it by the number of stations that share the network. For example, for the National Bank to have a fully-loaded Ethernet segment that has 200 users contending for the same 10-Mbps channel, it delivers only 10 Mbps/200 = 0.05 Mbps of average throughput per station. This assumes that all stations generate the same traffic all the time, which in reality is never the case. In addition, this assumes 100 percent efficiency and a 100 percent loading capability, neither of which applies to Ethernet. Due to the chaotic nature of shared Ethernet networks, the average utilization is limited to less than 40 percent, which means that the average throughput rate is limited to about 4 Mbps for a large Ethernet network. For switched networks, the average throughput can approach wire speed, since a switched network does not share the random nature of a shared network.
The National Bank also possesses a Token Ring network. To improve the performance of this topology the first option is to upgrade existing users with Token Ring switches. Token Ring is a shared-media technology, just like Ethernet. Switched Token Ring uses the existing NICs and requires a new switching hub that provides a dedicated connection, allowing for dedicated 16 Mbps of bandwidth for each node. Token Ring switches are relatively new; however, the huge potential of this market is slowly being realized and many firms, including the National Bank, we recommend migrate to Token-Ring switches.
Ease of Migration:
To assure an easy migration, the analyst or network manager must consider at least five questions:
Replacing equipment is always disruptive and time-consuming and should be avoided wherever possible. For example, replacing network adapters should be avoided at all cost, while the cost of a new NIC alone may not seem significantly high; the cost of installation, configuration, and associated user disruption often exceeds the cost of the NIC itself. If replacing a single hub, instead, can increase the performance of the network, it is recommended that the network administrator choose that route. Replacing equipment prematurely should also be avoided. Networking gear is part of The National Bank’s capital budget, meaning that the equipment needs to last for a period of, at least, five years before it is in effect paid for.
When new users are added to the network, it is suggested that the network manager choose the best available equipment at the time. These new users will have faster machines, requiring a higher-speed network connection. New users and servers will need to be integrated transparently into the existing infrastructure.
Another point to consider when upgrading to a new high-speed LAN is what to do with the old equipment. Often new high-speed equipment is added one step at a time, replacing at least some existing equipment. The cost analysis needs to factor in if the replaced equipment can be used somewhere else or if it becomes obsolete.
The large-scale upgrade of the existing 486 processor workstations to Pentium 100 with 64 megabyte RAM and 2.1 gigabyte hard drives will occur over the long-term for the 2500 PCs. The vision is for a 4-year period, in which an abrupt cut-over strategy will be employed. This conversion strategy is necessary due to performance counterindications, since the existing system may significantly hinder performance due to the fact that the older PCs are obsolete. Thus it makes little sense to explore a parallel conversion strategy, since this may further increase the cost of the overall system/network migration process.
Data
Conversion:
The above figure features a data conversion model. One part of preparation required for implementation is data conversion. Seeing that the old system contains a flat-file structure, the need to convert this to a database approach is evident.
The process of this conversion requires the help of many components that play their part in transferring flat-files to database tables. The components specifically are:
Hardware Conversion Details:
Architecture |
Speed |
Medium |
Type |
# of Routers |
Size |
| Standard Ethernet | 10 Mbps (56 Kbps avg. throughput) |
Standard cable | Semi-private |
16+ |
Large |
| Token Ring | 36 Kbps avg. throughput | Enhanced cable | Private |
8 |
Small |
| FDDI Ring | 50 Mbps | Fibre | -- |
-- |
-- |
Hardware Conversion Costs:
| Item | Cost |
| Router | $8000 each |
| Leased line | $1000 each |
| FDDI hub | $500-$2000 per port |
| Conversion from 486 processor to Pentium 100, 64 Mb RAM, 2.1 Gb HD | $6,000,000 for 2500 PCs |
Information systems personnel must come up with a solution for
moving data between incompatible systems in a reasonable timeline and assure that data
integrity controls for the accuracy, completeness and timeliness (or synchronization) is
properly maintained.
The risks to system conversion are enormous and include:
- Plummeting productivity during and after the conversion period
- Time and cost overruns because of unexpected problems and last-minute surprises
- Discontented bank customers and employees (if things
go terribly wrong)
- Permanent loss of access to critical information and even to the documents themselves
- Hidden problems that continue to crop up and linger long after the conversion is over
The criteria for success include reduced risk, acceptable cost levels, a reasonable
elapsed time for the data conversion process, easy setup and maintenance of data files,
databases and backup/archival storage, and simplified migration of legacy systems data to
and from the new enterprise applications. Interim solutions may be necessary, such as
establishing bridges to transfer data from the system in the process of decommissioning to
the other bank branche's data center, which will now absorb a greater workload.
System migration or conversions involve a strong project management approach in order to
accomplish the objectives of merging or transferring data from one entity into the new
platform. There are many issues involved, including negotiating revised licensing
agreements with the successful vendor, as well as consummating new maintenance contracts
and preparing Request for Proposals for additional hardware or support products. The
administrative processes involve the following tasks:
- Assessment of user needs and the criticality of applications
- Impact analysis, scoping and constraints/dependencies, including the degree of binding
between bank applications
- Staffing the business and technical teams
- Strategic planning for technical and business roll-out (to coincide with branch openings
and closures)
- System selection and cost/performance analysis
- Project management and scheduling
- Methodologies and modeling techniques and tools
- User training and acceptance
- Computer security issues (including implementation of internal controls)
- Support and Help Desk
- Common terminology and standards
The technical issues are also challenging. One idea that may captivate IS management in
both the bank and exchange bureau level is to consider a whole new solution, such as
client/server system. However, the luxury of time would preclude this option for most
cases. On the other hand, network configurations may preclude an easy, immediate solution.
However, "open systems" architecture is one of the directions that the new bank
might be heading, but the term is one of the most widely used, but least understood IT
buzzwords. In computing, open systems seemingly implies compliance with standards,
industry adoption of those standards and wide customer acceptance of technology based on
those standards. But, open systems actually have nothing specifically to do with
computers. "Open systems" is simply a market dynamic that enables banks and
companies to build mature commodity industries. Because technical design and
implementation is not done in a vacuum, IT must listen to the bankers and conform to the
overall merits for merger.
The technical issues involve a panoply of topics, including:
- Hardware and Software Installation
- Upgrades to Core Bank Applications and Compatibility Issues
- Transfer and Reorganization of Databases and Master Files
- Porting Interfaces
- Inventory, Cataloging and Indexing of Software Modules
- Prioritization of Bank Applications
- Timings, Performance Tuning and Capacity Planning Studies
- Use of Automated Tools (i.e., CASE)
- Testing – Test Plan and Scripts, Baseline, Regression, Metrics, Problem Resolution
- Milestones and Certification
- "Data scrubbing" Techniques and Translation
- Availability and Access to Current and Historical Data
- Change Management for Application Software and for the Transfer of Segments of Data from
Large Volumes to the New System.
Enterprise data infrastructure tools may be necessary to handle special database problems.
One of the dangers of "manual " application conversion projects is that they
often result in a badly denormalized database. This alone will spell trouble for the
"happy" customers already residing on the existing system. The manual approach,
in short, does not provide a stable foundation for the future use of SQL based tool, or
database enhancements. For example, the project team managing a general ledger or bank
system conversion would need to devise or create a data conversion plan, an application
test plan and a detailed task plan. This effort should include the documentation of
current system interfaces and data flows. Additional responsibilities include: data
testing, data mapping, process flow documentation and the coordination of affiliate
conversion and data migration.
Some of the following technical tasks that need to be addressed are, as follows:
- Populate and refresh databases
- Build testing environments
- Data transformation
- Migration of large production databases
- Create data bridge programs for reengineered or new applications
- Customized design and implementation of new modules ("to fill in the gaps")
- Support major migration to new architectures such as client/server or more distributed.
Major migration plans should be customer-related and revenue-generator type applications.
The business functions are more important than the actual movement of data to the new
environment. That is, the mapping will be mechanical, but the business rules surrounding
the use of the data has dire consequences. What happens to a customer’s "bounce
protection" if the linked savings account is not moved over concurrently or if the
customer is relying on a credit card advance which is weeks or months away from
conversion?
Migration or conversion problems are many. They may involve organizational changes,
including a tiered management structure with system administrators on one side and
business experts on the other. The methodologies used help create a framework for analysis
and design of the bank applications, and modeling assists in the implementation of the
degisn and links the data structure to the business rules. The business plan will dictate
what branches are to close or to be retained in the network. As a result, branches will
undergo renovations, including a new teller platform, LAN/WAN interfaces, manuals and
documentation, forms and signs. The need for careful coordination is essential to assure
the proper transfer of the customer base into the new system in the phases in the project
schedule. This is not an overnight event. Also, training and tool evaluations are areas
that can hamper a smooth transition. One tactic that should raise the eyebrows of bank
examiners is the "fix-on-failure" technique. Finding bugs or problems after they
occur rather than before implementation may result in extensive reprocessing of the data
as well as the generation of complaints.
The success criteria for migrating data over to the new platform can be measured beyond
customer satisfaction. Some benefits include:
- Predictable costs, schedules, and results (actual versus plan)
- Expected translation of essential features
- Minimal down-time during the process
- Few or no irreversible mistakes
- Employee competence in the required new skills and behaviors.
Conclusion:
Bank examiners should be concerned about big merger projects. An opinion on whether a
conversion project can be deemed "satisfactory" is subjective; nevertheless, the
implementation and support issues described above may help to formulate such an opinion.
It is easy to blame project failures on incompetent staff, uncooperative users, unsuitable
commercial software, and insufficient planning. The real reason for project failures,
which entail schedule delays, budget overruns and poor quality, involve the nature of big
projects. These factors include: (1) the consequences of a major mistake buried or
overlooked; (2) internal communications problems; (3) the demanding and changing business
conditions and requirements; (4) the sense of frustration that engulf groups, and (5)
staff turnover.
| Banking System | Expected Results on a Peak Day |
|---|---|
|
125 trans. per second (TPS) during peak 4-hour period CPU utilization: about 90% Running on an Enterprise Server Initial determination of data overhead for AIX and DB2/MS-Access is 12% |
Location Decomposition Diagram:
Location Connectivity Diagram:
The National Bank contains 640 branches broken down into the following regions: 432 Quebec, 119 Ontario, 38 Maritimes, 9 West Canada, 45 New Concept Branches. The new concept branches feature one-stop banking (e.g., Maxi & Cie, Bureau de Poste, Caisse Populaire). The location connectivity diagram presents one network geographical structure broken down into the folloiwng regions: 19 Quebec, 2 West Canada, 4 Ontarion, 1 Maritimes. All network geography regions are connceted via FDDI.
Implementation Schedule (Gantt Chart)
