Business Daily from THE HINDU group of publications
Monday, Mar 26, 2007
ePaper


eWorld
Features
Stocks
Cross Currency
Shipping
Archives
Google

Group Sites

eWorld - Interview
Info-Tech - Software
Web Extras - Outsourcing
`Sharpen your data models'

D. Murali

David Dichmann of Sybase shares insights with eWorld.

The fight against hunger is an ongoing one, worldwide. The United Nations launched the World Food Programme (WFP) in 1963 to feed the needy, such as refugees, internally displaced people, and school children. WFP feeds nearly 10 crore people in about 80 countries. "Initially, the distribution and tracking functions were managed out of a central office using a flat file FoxPro database with a DOS user interface," reads a description of the WFP-supported Ecuadorian School Feeding Program.

"Provincial coordinators would call into the central office and activities were managed by telephone; four data entry people updated the information. The application did not have input validation. Predictably, errors crept into the system. School eligibility records were frequently duplicated so some schools received double orders while others received none." Thus narrates www.sybase.com, the site of Sybase, in a section on `Customer Success Story'.

Sybase provided a Web-based solution to WFP, using PowerDesigner, to manage nationwide, decentralised food distribution and ensure that 22 provinces have access to the solution for monitoring and updating their school eligibility and participation information. "The efficiency of the new system saved approximately $1 million in operational costs out of last year's $25 million budget," says Robert Pazmiņo, systems manager of the Ecuadorian School Feeding Program, cited on the site.

"The annual saving of $1 million is equivalent to feeding an additional 60,000 children for one year," says David Dichmann, Senior Product Manager for Design Tools at Sybase Inc, in an interaction with eWorld. "The success of the project, as in any technology situation, is due to the right models," he insists. "Data modelling is a critical skill for IT professionals including database administrators (DBAs), data modellers, business analysts and software developers. It is an essential ingredient of nearly all IT projects. Without a data model there is no blueprint for the design of the database."

Dichmann manages the technical marketing, vision and direction of PowerDesigner, `a leading analysis and design tool for use in business information systems development'. He has over 16 years industry experience in both technical and business roles, working with small, start-up, and established businesses. Here's his take on a few questions.

What are the latest ideas and trends in modelling?

The essential principles of data modelling are well understood. We have been doing this for some time now, and all understand the value of creating physical data models to understand a database structure. The growth and new value today comes from going beyond the physical into structures around data in motion - business intelligence, data warehouses, data federation and movement - all encourage more conceptual data analytics. Data modelling is developing more into enterprise information architectures, and this is relating to more general enterprise architecture, as a way to provide clarity from the complexity of modern information systems. The value of this is clear when we look at the problem in a practical way.

For example?

Say, we want to provide a 360-degree customer view. We need to know where `customer' is used throughout our systems. At Sybase, we examined our internal systems and found several hundred representations of `customer' throughout our databases and systems. With data modelling, we can start to define and describe each instance. But, bringing data modelling to the next level, we need to define the dependencies and mappings between them to get a comprehensive `where used, what's the source,' type of detail. Tools can bring new innovations to this problem, such as our unique data mapping editor for easy drag and drop dependency mapping.

A common problem?

Many of our medium to large customers have similar problems - like how to provide comprehensive 360-degree customer view, or to defend regulatory compliance around financial reports, or to gain better agility in business software development. Measures like quality, integrity, and security matter a lot. Modelling techniques can provide a method to easily capture this information. Once we have a clear picture, or model, of the data environment, and if we do that in context with the application and business environment, we can more easily use techniques such as integration, federation and movement to solve business analytics problems, or provide better customer self service, or to run a more competitive organisation.

By managing the blueprints of both the data and application in an integrated way, and aligning that with the overall blueprints of the business (business processes, business goals, etc) we can more easily track information sources (or heritage), measure impact analysis (the impact of a change throughout an enterprise) or achieve agility through business and IT alignment.

Virtual environment can provide access - for example, a customer may want to see all his accounts on one Web screen - define and describe the source and the integration, federation, warehouse, etc.

What are the levels of sophistication required in India?

India is no different in the level of complexity - not the amount of the information but the demands to use this information in new and different ways. In a global economy we need to defend the quality and integrity of reports. India is no exception and this requires a way to understand that is approachable and reliable and modelling provides that abstraction.

A modelling message, to wrap up.

Sybase sees modelling as a top shelf, strategic initiative; it is integral to the unwired enterprise.

The goal is to make the modelling technology approachable.

The integration of all the modelling makes this complete. The need to centralise information and securely manage multiple systems through acquisitions, mergers, or regulatory compliance - this creates a catalyst to move into more and more need for modelling and more and more need for metadata management.

In opportunistic development, where there are cowboys on keyboards, the problem is that the front line developers do not see the value of modelling. They do not see the bigger picture and what we find is that many coders do not see the value of the code being shared at a broader picture. But at the higher level, we see there is greater savings in reuse.

Relevance to back office, service centres and so forth?

Outsource and offshore information is coming from many companies in many countries. Let us consider a customer service system. When a customer makes a call to the customer service centre they expect a quick response. And it has to be as quick as possible, but when organisations are outsourcing this work, and the customer's information is coming from different countries, we really have a need to be able to consolidate in a timely manner. Using techniques to integrate and federate data, we can be sure that when the call comes in, that the aggregation happens, and is accurate and reliable information.

Is that already happening?

Yes. Technologies such asSybase Federation, acquired when we purchased Avaki, are giving customers this option. What this technology does is create a virtual representation of multiple source databases, and from this concept it will define mappings. For example, we can take a column from the customer table and analyse the information at the call centre, but the original information is retrieved following a predefined path to where the information is actually stored. However, this is just one method of achieving this consolidated view. If we need to ensure a different level of integrity or timeliness, we may also consider data movement technologies such as replication, or process-driven integration technologies, with search and security layers. Depending on the needs of the ultimate user of the information, the infrastructure being used to manage the information and the volume of information being accessed, one technique may be better than another, but in the end all techniques will be used together in some way. The goal is ultimately allowing the enterprise to create a representation of action-ready information that can then be made available to any terminal, computer or device anywhere in the world. This is Sybase's vision of the `Unwired Enterprise'.

What about security?

Absolutely, security is definitely required. Sybase has a strong history considering our presence in the financial industry. Sybase technologies are used in all of the world's banking sectors, Wall Street, American banks, German banks, French banks and more.

Any Indian banks?

Many that we are currently working with. Such as HDFC Bank and Indian Overseas Bank.

Current challenges?

Looking from the modelling tools perspective, `define' and `describe' are the challenges. Most organisations have basic shell knowledge of what their information is doing. In order to get to the virtualised store, you really need to reach a higher level of define and describe that is really accurate. For instance, when we change the way we do business, and we change the process, we have to also change customer definition and related screens.

It comes down to models again.

Yes - it comes down to models again. You can build what you have and document and reverse engineer and `auto document' based on server knowledge. If you do not know what you have in the source, you cannot apply new and creative ways to access. If you do not know what you have, it is very difficult to leverage technologies like Avaki or IQ. Therefore, IT development and especially IT managers need to be more in line with the business, and that also means more in line with the customer. The managers need to have a higher level of understanding of the value of modelling overall, to know what the model can do. But the effort can be evolutionary instead of revolutionary. Taking the customer consolidation example above, just knowing that there are 350 instances is a great start, and as we begin the efforts of consolidation, federation, integration and so forth, we can evolve the knowledge base over time.

Are we giving enough attention to data modelling?

We've given a lot of attention to the traditional data modelling - that I have a database and I know what I can do with it, very narrow scope - but what is happening is things like regulatory compliance force us to roll this to the enterprise level - we need to have this to get better support to our customer, supplier, compete at global level. Not enough attention to data modelling, I'd say.

On the learning of modelling.

That depends on your subject matter knowledge and experience with modelling. For example, if you are a data professional and understand the concepts of third normal form, then the modelling tool will be relatively natural for you. PowerDesigner was specifically designed for that purpose, using intuitive user interface techniques and a modern, highly graphical Windows interface. However, with limited modelling knowledge, we find that you can learn the basics in a matter of a few days of a training class or you can self-study through books in the market. It is not necessary to have a computer science degree. However, we are finding many colleges, including in India, are teaching the principals of modelling and metadata management as part of any modern systems architecture for information technology.

How are you planning to take modelling to a wider level?

Sybase has done an extensive study on modelling and enterprise architecture. When it comes to broadening the scope, the techniques themselves are not so new. PowerDesigner has capabilities within it to allow us to create the higher level diagrams, like conceptual data models or analysis-level UML models, and generate lower levels of specification from them. After this generation, PowerDesigner automatically remembers the heritage. Having a practical and applied toolset is the key - we want the natural effort of modelling for better design to be an asset to the development lifecycle, not an added burden to the developer or the DBA.

When the technology changes, what happens to the model?

The models will adapt to the changes made in the enterprise. In actual fact, if we have the models, or the blueprints, of the enterprise, we can easily measure the impact of change. Changing the technology may have an impact on the business processes, and needs to be communicated using terms the business understands. PowerDesigner's synchronised models allow for that communication to happen quickly and efficiently. Once the technology change is made, we can easily adapt either by making the change through the model and generating it back out to code, or by reverse engineering code into the tool to ensure the models match the outside world. Using these techniques, all changes will be managed and updated naturally.

How you are making models future-proof?

Current consensus - customer survey, media and analysis state 60 per cent of projects in corporate IT fail. This is not because we develop bad code or databases; it is because halfway through funding is cut or project is finished but not used. It is because we missed the goals of the business. If we can make projects more agile - so we can change the project to meet changing business goals or changed interpretation of business goals...

On your India operations.

Sybase has a development office in Pune. Engineers in India are on par with the other development centres worldwide. Pune office is responsible for critical elements of many different software projects we have throughout the company.

On the relevance of modelling to outsourcing companies.

When you have an outsourcing project, modelling and a clear sense of business and technology architecture are needed. The detailed specification becomes very important. The communication between those who are doing the project and those who require it should be much clearer. Outsourcing is different from off shoring. When organisations move development `off shore' to India the development is still owned by the parent. In outsourcing we find that the responsibility to challenge ambiguity in the specifications is less. This is where we really need modelling to ensure the development is in line with what the customer (the organisation paying for outsourced development) wants.

So the communication between the customer and the outsourcing organisation must be much more clear and precise. Outsourcing organisations are not tasked with the need to judge whether the product is good or not, whether the product is going to help the company or not. We can solve that communication gap again using the model. We can reduce the mistakes by understanding up front what is needed, before we develop something wrong. The issue is in the way the outsourcing is managed. In outsourcing the company is paid to develop to a spec. In off shoring the company owns the development team directly - so there is a greater responsibility to the company to correct misunderstandings in the spec. In outsourcing there is less of a burden from the firm receiving to make corrections so we have to be clearer in the requirements and have them very well modelled in a clean, understandable way - then the results of the outsourcing development will be more in line with what we want. This removes the risk of developing a perfectly reliable, and perfectly useful environment but not what I wanted.

Any type of outsourcing produces this - because we are paid to produce a result to spec. I am paid to paint a room, so I ask for orange, you paint it orange, and never think to ask me if I am sure I like this colour. If I see the result and do not like it, simple, pay again and we paint it a different colour. But in off shoring - since we are all the same company, we are encouraged to question more - so we may ask, "do you want orange" - and get some deeper understanding anywhere there seems to be ambiguity. We will pay 2/3/4 times to get it right if it is not our home we are painting. But now with competition in outsourcing, getting this connection becomes more important. If we can use models to close this communication and get the feedback sooner, faster, we can get things right - nobody wants to build the wrong product.

Do specialist groups work only on modelling?

That depends on the level of modelling. Modelling is not development. Modelling is part of development, it participates in the development lifecycle, but it has more in it than just a faster way to write code. Modelling is the requirements, the analysis and use cases, and the overall agreement between business and IT for the goals of the whole enterprise architecture, at a project level or at the top level. Then, design models, derived fro the analysis and then refined by technical staff like DBAs and developers, will represent the actual code. This ensures that the gap between the conceptual framework (that we want to have developed at a business level) and the implementation (the final code, database and system) is in line.

Who are your competitors for modelling tools?

Take a look at modelling tools on the market: there're modelling tools for business, for developers (UML tools) and there are data modelling tools. We are #1 for DM tools (MQ and Market Share) compared to competitors Erwin and EMBT. For UML we are competing with IBM/Rational and the others there, and for BPM there is a large mix. Rather than create silos, which are not related to each other and to data, we put all these into one tool, and put all the techniques into one tool - but we are the only ones that have related all this together to the final implementation for breadth and depth. Other tools give breadth or depth but not together. Like AutoCAD - draw the image and get wireframe, etc, we get the same thing with software, the business view in synch with data, application and process so all levels and layers are in context. In this context we really have no competition.

What about acquiring India companies?

It is about fitting the technology to fit our strategy - there is no reason to exclude any one country from another. If there was a technology we need and the right company in India was a fit, I do not see us hesitating to make this acquisition.

MuraliDe@gmail.com

More Stories on : Interview | Software | Outsourcing

Article E-Mail :: Comment :: Syndication :: Printer Friendly Page



Stories in this Section
Scoring more hits


Clicks and chicks
IE error message
Setting `details' view
Hand IT to the expert
Managing data the smart way
`Sharpen your data models'
Quiz
A practitioner's manual
Cartoon


The Hindu Group: Home | About Us | Copyright | Archives | Contacts | Subscription
Group Sites: The Hindu | The Hindu ePaper | Business Line | Business Line ePaper | Sportstar | Frontline | The Hindu eBooks | The Hindu Images | Home |

Copyright Š 2007, The Hindu Business Line. Republication or redissemination of the contents of this screen are expressly prohibited without the written consent of The Hindu Business Line