Contact Editor John Dix

By John Dix
IntraNet, 5/19/97

No one can accuse Fidelity Investments of thinking small when it comes to intranets. The mutual fund powerhouse expects its20,000 employees to be using Web browsers to access the bulk of its information resources within three years.

While this will vastly simplify the task of managing desktops in this worldwide corporation, even more important is a complementary effort to pull the firm's mammoth data warehouse into the intranet fold.

Together, the efforts represent nothing short of a fundamental transformation of the way in which Fidelity distributes and uses mission-critical information, says Timothy Davis, director and principal technologist of the company's Administrative Information Management Systems' Data Warehouse Group.

So much for posting dental forms.

Actually, Fidelity already uses its intranet to support these more mundane functions. But the larger goal is to influence the very way the company conducts business.

Data delivery

"The cost per desktop of the old paradigm, a Windows PC with lots of applications, has become quite significant,'' he says. "And even more significant is the maintenance of the desktop. Anyone can download software so every desktop is unique. We get into a lot of conflicting applications and a lot of desktop configuration issues. So we're heading toward more of an Internet delivery mechanism with thin clients.''

The typical Fidelity desktop today is a Windows 95 or Windows NT machine outfitted with Microsoft Corp.'s Office, a Netscape Communications Corp. browser and a host of other applications specific to the user's needs.

"Most of the problems are with stability. We have conflicting software, conflicting drivers and even problems with database connectivity," Davis says. People pull their hair out around here because of these things. Moving to an Internet desktop would solve that be-cause your connectivity is now on the back end as opposed to the front end.''

Davis says the Internet desktop vision has browsers serving as universal front ends to all other applications, even Windows applications. Applica-tions become modular and network-based. Users access only those application functions they need, cobbling together custom programs on the fly. "I may end up with an application that's 10% Word, 2% Lotus Notes, 5% this, 5% that,'' Davis says.

The world of component software won't arrive overnight. "It's probably two to three years out,'' Davis says. "Java is still developing, the whole language base is still in flux, and we have a lot of users tied to Windows. So it is going to be a progression. But as the functionality and capability of the browser improves and more importantly, the network bandwidth issues start to go away, this will become a much more effective delivery vehicle.''

Asked whether the current industry focus on cost-of-ownership issues is driving the company down this path, Davis says it is more a matter of survival: "We have a certain head count and a certain set of resources to deliver the information we need to deliver. If we don't become more efficient at what we do, we aren't going to meet our milestones.''

Data stockpile

The principal challenge is providing a uniform view into Fidelity's varied systems, which include multiple trading systems, accounting systems, custody systems and pricing systems, many of which are scattered around the world.

"It's been very difficult to get a clear, consistent view of worldwide information, worldwide trading, worldwide holdings, and most important, worldwide exposures,'' Davis says.

Fidelity is pursuing a solution based on a data warehouse from Red Brick Systems, Inc. and a host of decision- support tools. The latter make it possible to pluck information from the warehouse and other resources and deliver it to different types of clients, be they Web browsers, Windows-based machines or pagers.

The decision-support tools are the glue in the middle, a middleware layer that couples the warehouse to the desktop, resulting in a three-tier architecture.

The top layer is the Fund Intellect data warehouse, an SQL system that contains data gleaned from Sybase, Inc. and Oracle Corp. databases, as well as other data sources. This warehouse was originally developed to give trade controllers a view into trading activity -total trade volumes, where trades stand in the settlement process, what the life cycle of a trade looks like and historical information going back seven years.

"A trade goes through many steps in a process,'' Davis explains. "Once it's executed, it can go through a cancel process, a rebook process or even fail outright, which represents an exposure for the company. So it's very important for us to track and understand which brokers are giving us good executions and which markets we are going into.''

But SQL constrains how the warehouse can be used. "From a business standpoint, SQL is a very limited language,'' Davis says. "Typically, you want to ask questions like, 'Show me this month vs. last month,' but there is no structure within SQL to do that. So you build temporary tables or do multiple passes and try to pull all the data back to your desktop. But if you have a 200G-byte warehouse like we do, it becomes prohibitive to do that. The whole reporting system just breaks down.''

Enter Fidelity's decision-support system, which it calls Trade View. Trade View is actually a collection of information tools Fidelity uses to help make sense of the reams of data the company collects and generates. The idea is to run these tools on a Unix server dedicated to the task of querying the warehouse, building intermediate data sets, and then converting the results into reports that can be viewed in various ways.

Together, the tools represent "an active middleware that basically moves the processing from the fat client desktop out onto a server,'' Davis says. "The desktop simply becomes a dumb device, and all the processing occurs out on the network.''

The various decision-support tools all fit into a framework provided by Infor-mation Advantage, Inc.'s DecisionSuite product. DecisionSuite lets users build queries using business terms as well

as look at things dimensionally, and it supports both browser and Windows clients, all of which are important cri-teria for Davis.

DecisionSuite also has two other advantages: a meta data focus, meta data being information about the data in the system; and support for proactivity, meaning it can alert users when certain conditions arise.

"Meta data is really the connection piece with the users,'' Davis says. "It

lets them access structured and un-structured data and find information about things like, 'When was this last refreshed,' and 'How often does this show up?'"

Users can build reports by simply dragging and dropping elements. Users could, for example, include a time period, a trade and a portfolio, and specify a trading desk, a currency, country, market exchange and so on. "So they're working in an environment that's familiar to them," Davis says. "And, through the meta data, the Information Advantage product then translates that into queries. It may have to do a couple of hundred queries to actually build this one report."

The resultant report can be sliced and diced into more granular reports, all with different criteria. Fidelity sometimes calls these reports "data cubes" because they enable users to look at information in lots of different ways.

DecisionSuite also allows security to be brought into play, so the fixed income trading desk only sees fixed income trades and doesn't see equity trades, and vice versa, while trade controllers can see all trades. "This is very important because it provides a common security base across the Windows desktops and the browsers,'' Davis says.

Proactivity is provided through DecisionSuite intelligent agents, which Davis describes as human surrogates. "A user may have several hundred agents looking for different things, and when one of them finds something, it sends out a notification. That's important because few people have three or four hours to spend querying their way through the data warehouse. We have to build these systems so they are proactive, so they're alerting people.''

In practice, this will enable risk managers, for example, to use agents to look for exposures, such as limits being exceeded, or trade controllers to identify unusually large trades. And DecisionSuite can deliver the alerts to Windows desktops, Web browsers, pagers or even mail systems. This specifically addresses Davis' goal of increasing efficiency.

Global perspective

is no small problem. The data warehouse is up and running, but the Information Advantage products are only being used by 20 people in Boston and another five or six in Texas. The plan is to roll out the products to 100 people representing a variety of Fidelity disciplines.

But that raises the issue of whether to keep the warehouse and decision-support tools central or distribute them around the world. "We're making a stab at keeping it central, having people in foreign lands directly access the warehouse over the company intranet,'' Davis says.

While there is some question whether network delays to places

such as London and Tokyo will be a problem long term, the impetus to keep it centralized is huge. The company does trading worldwide, so there really isn't an end of day. The warehouse is almost always humming.

"When a trade comes in, we do load, index and referential integrity checking, and try to maintain aggregation levels across the system in a continuously flowing process,'' Davis says. "These are not trivial tasks, and you have to have it very well synchronized for it to run effectively.''

Further driving the desire to keep it central is the company's move to trickle feeding the warehouse. While many companies update their warehouses once a month or once a week, Fidelity wants to update its warehouse every 15 minutes. "That will enable us to do intra-day decision making,'' Davis says.

Sound ambitious? Everything related to Fidelity's intranet-centric vision of computing is ambitious. But the highly competitive world of mutual funds demands nothing short of full throttle. After all, when information is the lifeblood of the industry you compete in, you can't afford to be left behind.

"The intranet is going to make delivery of data very easy, but it doesn't solve your data quality problem," Davis says. "You have to put your data warehouse on the intranet and set up the infrastructure so you can integrate your structured and unstructured data."