ATG Data Administration / FOBOCA

CLIENT: CSFB-Credit Suisse First Boston, Eleven Madison Avenue, New York, NY

Credit Suisse First Boston is a leading Global Investment Bank. Corporate business information consists of more than 200 database models, covering all aspects of the banks activities and supporting infrastructure. A global Data Administration/Data Modeling team provides standardized design for Meta Data across the enterprise. This approach prevents redundancy and facilitates integration, transparency and re-use for Users and Applications.  Due to the extreme monetary responsibilities involved this company places the highest standards (we have seen so far) for Meta Data design quality and documentation.

ANALYSIS/OBJECTIVE:

As part of the team, Jayzed Data Models designed 10 new databases and implemented changes on 60 existing data models.

SOLUTION:

This project is still in process. Utilizing prior experience, we introduced efficient data mining techniques shared between backend queries and PC based tools to reduce human workload and provide higher analysis quality.

Product Support Database

CLIENT: REUTERS/Sailfish, 44 Wall Street, New York, NY

REUTERS known worldwide as a newswire service and provider of real-time financial market data. Their subdivision 'Sailfish' is  a market leader for Risk Management Software featuring KVAR+® product line. The system computes the total Value at Risk from an extensive set of derivative pricing functions the portfolio and the set of market rates, respectively rate series.

ANALYSIS/OBJECTIVE:

The production database for such a risk management system has considerable size: approximately. 150 tables, 200 stored procedures and 300 triggers. The product support team employs various semi automatic proprietary testing tools and methods for quality assurance. There were two requirements to increase the level of automation - thus providing cost savings:

1. Reconciliation of test scenario and testcase data into a single open architecture.
2. Incorporation of Trigger Testing into this architecture.

SOLUTION:

Jayzed Data Models proposed and realized a Product Support Database. To achieve the first objective a Generalization approach was consequently applied in the Entity Relationship Model. For example a testcase is modeled as a Supertype entity having attributes for all generic data. Information further describing the specific testcase category like (in-, out- parameters, results) was modeled into a set of dependent Subtype entities. A Category Entity allows the selection among the subtype entities. As a result the reconciled data is still accessible for all operational systems - the test-tools. In addition automated aggregation for all tests is know possible for reporting and decision support. This concept is realized for test scenarios, testcases and regression tests.

Statistics

50 tables, 15 stored procedures, Sybase, ERWIN

Xentric Bonds Database

CLIENT: The German Stock Exchange, Frankfurt

There is probably no general information, I can add about this client. All electronic trading is handled by the exchange's XETRA® system. XETRA is a pure market system - it does not contain any historical data. The order book contains only open positions. The department Vendor Business Applications develops value adding products for institutional financial customers. Business Vendor Applications features the Xentric® product line.

ANALYSIS / OBJECTIVE:

The latest product Xentric Bonds® is now in production at major European banks. It provides fixed income security trading departments with a windows interface to XETRA based on a 3-tier architecture and of course it captures transient data. During the design phase it became clear that only a relational database could provide for the data access and storage requirements.

1. The database had to be designed and implemented in record time.
2. As a real-time Online Transaction Processing System it had to meet tight performance  criteria.
3. Development RDBMS is MS-SQLServer 6.5, production systems Oracle 8 and SQLServer

SOLUTION:

Brought into the project at this point, despite the tight schedule Jayzed Data Models employed a full database development life cycle. To address the top project risk the technical lead for NT networking and myself the expert for databases developed a prototype to prove the feasibility of the designated architecture: MS-SQLServer 6.5, Visual C++ database classes, ODBC and DCOM database server process. In close teamwork with the Stock Exchange's trader domain experts Jayzed Data Models established the Business Rules, Entities and Attributes. The resulting Entity Relationship Model complied to 3rd Normal Form. Using a proven reliable design tool - Silverrun the logical model was easily transformed to physical. The resulting schema script automatically generated and deployed on the engine. The assure the system performance and productivity or the development process Jayzed Data Models implemented all transaction logic in a stored procedure level. Thus avoiding code duplication, reducing the C++ part and   the Total Lines of Code. Finally the deployment on the production RDBMS proved just as easy.

Statistics

30 tables, 40 stored procedures, MS-SQLServer/ Oracle, Silverrun

AMANDA

Client: START Informatik, Frankfurt

The START group the biggest provider for Online Travel bookings in Europe. It's subsidiary START Ticket dominates the German market for Event ticket sales. This year's EXPO 2000 event with expected total of 50 million tickets is sold and booked exclusively by START Ticket's infrastructure

ANALYSIS / OBJECTIVE

A ticket transaction involves 3 parties:  Customer,   sales Agent and the event Provider. Furthermore there are 3 potential stages: Reservation, sale/Payment and possible Cancellation. The core accounting module computes the resulting periodic billings between Agent(s) and Provider(s) especially contributing all fees to the entitled party. The system was due for a complete redesign. The new system marketing name AMANDA had to ...

1. Ensure correct billings for all possible transaction cases.
2. Guarantee a preferment batch run.
3. Provide additional marketing statistics.

SOLUTION

Other billing systems try to include the statistics within the batch. As convenient as this may be on heavy data load it leads to crash all performance constrains. I ruled out any such approach. In accordance with the Product Manager we set the priority on billings alone, having any statistics to be generated separately. In production AMANDA proved to be fast enough, even for the EXPO 2000, to ensure the clients to bill on time. Statistics were scheduled on demand or between the billing batches.

In a further release, a micro Business Information Warehouse was implemented storing process data during the billings run for reuse in statistics. This finally gave the user a record performance on that part as well.  

Statistics

250 table database, Informix 7.2, Unix embedded SQL/C, Silverrun

WIOLA

Client: START Informatik, Frankfurt

For information about START see project AMANDA.

ANALYSIS / OBJECTIVE

The overall START Ticked system architecture was based solely on Unix. This had two drawbacks compared to PC/Windows based Client/Server systems:

1. The user simple has Windows PCs at his or her workplace and wants to have the Decision Support integrated with his or her other applications.
2. The development environment on Unix was strictly 3GL, leading to considerable lines of code and effort to realize an application's function points.

 Realizing these challenges START's management moved towards a PC Windows based Client/Server architecture, outsourcing operational booking system to Management Data and evaluating a Data Warehouse Architecture for the Accounting and Decision Support modules. Having the expertise in Windows Client/Server development, I was asked to develop a Business Pilot DSS based on AMANDA - marketing name WIOLA.                                                                                                                               

SOLUTION:                                                                                                                                        

Where a full scale Data Warehouse would reconcile all operational data and replicate it on a separate database, a pilot in this case demonstrates the benefits but still accesses the operational system. The server architecture was pretty much set. The Informix database was fairly normalized and Business Information already rolled up in the previous AMANDA development. So the focus on this project is to evaluate, choose and implement the Windows BIW access application. After evaluation of Business Objects®, Cognos® and MS-Access®, I choused and personally recommend MS-Access. Technically the pure BIW applications lack the ability to customize a set of predefined inquiries. MS-Access comes with Visual Basic full scale programming language allowing to set up security, forms and procedural elements. Furthermore the users do already have MS-Office licensed, whereas otherwise considerable additional fees are mounting for every installation.
The ODBC connection proved to be reliable and easy to setup. The prototype was implemented in roughly a third of conventional development time.

Statistics

Informix 7.2, MS-Access, Visual Basic, ODBC