(Sun Microsystems, Inc.)(1997).iso/products/FrontPage/petris/petris1.gif)
Click here for larger view.
When Joe Frazier and Mike Frank decided to build an on-line store, they decided to build it big. Very big. By the time they get done completely stocking the virtual aisles of PetroInfo, Frazier and Frank hope to give their Internet shoppers access to over one million unique data files that energy companies use in their search for natural resources. Such a huge selection will practically guarantee that PetroInfo has what its shoppers are looking for, but by offering one million items for sale, the two men run the risk of turning a customers' search for valuable data into a search for a needle in a haystack.
Of course, searching for a needle in a haystack is exactly what PetroInfo customers are trying to avoid in their quest for natural resources. Oil companies, especially, incur huge costs in their search for oil. The price of drilling a shallow, on shore well might be a relatively cheap $100,000, but a deep water, offshore project could run up to $1,000,000,000. Costs like that make it imperative that each potential well site be chosen and exploited judiciously. To take the guesswork out of their search, energy companies turn to firms such as Petris Technology Inc., the Houston-based firm that opened the PetroInfo site in January [ed. 1/97] at http://www.petris.com.
Petris refines its customers' search for oil by capturing and analyzing well logs taken from freshly drilled, virgin wells. These logs measure and record various rock properties that determine whether a given well will yield oil, whether it is a "dry hole" or a "producer." The dry holes are capped to save any further, unnecessary costs while the producers are prepared for oil extraction. Once it has helped decide the fate of the well from which it was taken, a well log maintains its value because it can later be correlated with other data to aid the exploration of adjacent areas.
The search engine originally designed to serve
Petris Technology's PetroInfo site simplified searches with a
graphical user interface and drop down boxes. However, users who were
unfamiliar with the site would have had no way of exploring the data
available to them before executing a search query.
"Whether the well itself is dry or a keeper, the data in that log is valuable," said Frazier, production manager for Petris. "As the industry develops new technologies that identify or produce new hydrocarbons or, oil companies return to old, mature fields. With new techniques like horizontal drilling and 3D seismic coming to the table, all the horizons are being re-explored many times. Well logs are used to map all the elements that you know about. They're on-going data assets."
From the beginning, Petris assumed its on-line store would grow to huge proportions and identified the search engine as a critical component to its success. Shoppers cannot buy what they cannot find, nor are they likely to return to the site in the future. Perhaps most damaging, negative reviews of disgruntled shoppers would spread like wildfire in the energy industry and discourage new shoppers from visiting the site in the first place.
"We were really motivated to have people find the exact well logs that they need," said Frank, Petris's development manager. Consequently, text search engine such as Digital Equipment Corp.'s Alta Vista, Excite Inc.'s Excite, and the like were quickly ruled out because they could easily deliver useless results. "If you've ever searched for a bulk text string in Yahoo, you know that you tend to get back more information than you need and quite often you don't find what you need."
At this point, Frank decided to build his own search engine on top of an Oracle Corp. relational database. Using Web browsers, users specified parameters to use as search criteria, entered their values, and the database executed the search as an SQL query. Frank simplified the creation of search queries by building a graphical user interface that made extensive use of enumerated lists in pull down boxes. Deployed in an internal application that managed Petris's own relatively small collection of well logs, the database-based engine yielded more valuable results, but it could be "very slow if you throw in joins with wild cards," Frank admitted.
And the homegrown engine was not all that easy to use, even for Petris's own users who were trained to work with it.
"We could build a pretty complex query with a lot 'and/or' combinations for multiple data fields to refine a query, but a lot of users don't necessarily understand Boolean operators," said Frazier. "It's not as straightforward as we had anticipated it to be for the user building these complex queries."
Making it easy to perform fast and accurate searches on huge data sets remained an elusive goal until Frank attended a Netscape Communications Corp. developers' conference last spring [ed. spring '96]. There he saw a demonstration of Cadis Inc.'s Krakatoa Web Catalog Publisher, an application designed specifically for building search engines, or catalogs, for large databases such as the one Petris had in mind for PetroInfo.
The demonstration and subsequent investigation persuaded Frank and Frazier that Krakatoa addressed the major concerns they had making Petris's on-line store successful. The Krakatoa client software that shoppers use to search the PetroInfo site is written in Java, so anybody can run the application from any browser that includes a Java Virtual Machine, such as Netscape's Navigator, Oracle's PowerBrowser, or Sun's HotJava.
In addition to widespread access, Krakatoa establishes a straightforward user interface that resembles that of the Windows 3.x File Manager or the Windows 95 Explorer. The screen that greets shoppers is split in half. The left side displays a hierarchical tree structure -- a catalog of the data available at PetroInfo -- that expands and collapses to reveal more or less detail as users investigate various data categories and subcategories. As users navigate the tree on the left hand side of the screen, the right side of the screen displays the searchable parameters associated with any given level of the catalog. The experience is similar to flipping through the pages of a paper catalog.
PetroInfo ultimately opened for business using a search engine that lets users preview the site's data. As users browse through the site's catalog on the left side of their browser screen, they select the search parameters associated with any given section of the catalog appear on the right.
(Sun Microsystems, Inc.)(1997).iso/products/FrontPage/petris/petris2.gif)
On the server side, Krakatoa has effectively established a meta-data database of the Oracle database, which stores the PetroInfo data files. The advantage is not limited to the streamlined user interface. On the back end, Krakatoa pre-processes the search queries that would otherwise execute directly on the Oracle database. As a result, Frank has greatly reduced the time it takes to return a user's search results because the database server no longer bears the brunt of the query processing load.
"We have plans to scale up to millions of items and relational technology, especially for abstract or wild card searching, is not as fast as Krakatoa. Not even close," said Frank. "And our chances of getting hits are enormously higher because it gives the user an intuitive, logical way to drill into the data."
J.M. Huber is a perfect example of this data overload dilemma. The energy company's oil and gas division works with a number of different applications that contain data that is fed into a data warehouse and then analyzed using various decision support tools. The problem is that the applications' data comes from different sources -- production, accounting, financial, and other departments -- and the data is analyzed by a number of different people, including engineers, geologists, geophysicists, accountants, and managers. Not surprisingly, most of these users are not familiar with all of the applications from which they could draw data. As a result, they can easily overlook information that would otherwise prove helpful or even critical to their decision making. A catalog of information will ensure that all users know what data is at their disposal.
"What we want [Krakatoa] to do is deliver summary level data to senior management to identify a problem and pass that information down to another manager, who can then use the search engine to drill down and analyze it in detail," said Rene Calderon, vice president of J.M. Huber's oil and gas division in Houston. "For instance, if we know production is low in a certain field, we can search the field for wells that produce over 50, 100, or more barrels a day and also look at the accounting entries to find out what was going on in those wells during the same period of time."
As both PetroInfo and J.M. Huber's data warehouse demonstrate, an important part of data management in the energy industry is simply knowing what data you have and making it accessible. Cataloging data may well prove the most effective way to find those resource needles, whether they are hiding in haystacks made of data or dirt.