Year 2000 projects are giving businesses the "opportunity" to rethink and redesign their business-critical systems, whether they like it or not. Databases and the data they hold are central to this process. Unfortunately, it's harder to assess Y2K risk and test Y2K compliance in databases than in other vulnerable components (such as hardware or operating systems) that have fewer possible inputs and outputs and whose errors are more obvious. For example, it is easy to overlook the absence of one day's orders (say, those for Feb. 29, 2000--the year 2000 is a leap year) from a monthly rollup statement.
Moreover, a database system relies on the underlying hardware and operating systems for correct date and time information. It can be hard to determine whether the database is the source or the victim of a year 2000 error until a good deal of Y2K testing has already been done.
It's a standard disclaimer on vendor Web sites that a package is Y2K-ready only when it's operated and administered correctly and "provided that all products (for example, hardware, software and firmware) used with the product properly exchange accurate date data with it," as IBM states. PC Week Labs' database Y2K analysis highlights the importance of such warnings.
Any large organization is also likely to have several database servers that rely on each other for various tasks, such as user authentication, distributed transactions or data distribution.
Database data replication, for example, is a common way to distribute financial information from headquarters to all branch offices, but it's likely to involve many servers in several different time zones, possibly running on different hardware and using different operating systems.
The good news is that relational database systems, which are the most common database architecture today, gained prominence only in the 1980s, when storage wasn't so tight that years had to be stored as two-digit values. As a result, all the relational database systems now used store dates using four-digit years.
As extra security, the Entry Level SQL-92 standard specifies that years must be stored using four-digit values. All the big database players--IBM, Oracle, Sybase, Microsoft, Informix Software and Computer Associates International--have Entry Level SQL-92-compliant products, and have since the early years of this decade.
As a result, any organization that has been using the SQL date data type (also sometimes called datetime) is automatically storing all year information using four digits.
By comparison, fixing COBOL applications (which didn't have the same kind of standards that the SQL world developed) is a much harder problem. It would be sadly ironic if organizations found that their COBOL development costs have been contained in the short run by postponing migration to relational database platforms.
A particularly thorny problem is facing those using character or numeric fields, instead of real date fields, to store date and time information (as is common in older AS/400 applications). Should this be the case, those poor database managers should be canceling their 1999 holiday plans now--significant table design and SQL recoding is ahead.
Operational issues -- server administration and configuration tools, for example (where we found that problems handling the year 2000 leap day are still common)-- are going to be more trouble to handle. Unlike most century years, the year 2000 is a 400-year multiple and is therefore an exception to the normal "century exception."
Whether they like it or not, businesses will need to do some upgrading of their server software to address these sorts of problems, plus other vendor-specific problems, including the following.
Oracle7 and Oracle8 Oracle states that all versions of Oracle7 and Oracle8 are Y2K-ready. However, the default date format Oracle7 and Oracle8 use is to interpret all two-digit dates as dates from the 20th century. This is sure to catch the unwary by surprise.
For example, when we queried the last day of 01-Feb-00 in Oracle8 Version 8.0.4 using Oracle's LAST_DAY function, we got "28-feb-00" in return--correct for Feb. 28, 1900.
Both products provide a display mask (the letters "RR," rather than "YY") for data entry that provides a sliding window for two-digit years, which will handle dates from 1950 to 2049.
When we reconfigured the NLS_DATE_FORMAT parameter for Oracle to use "DD-MM-RR," the year 00 was treated as the year 2000 instead. We strongly recommend that companies using Oracle make this change in their Oracle initialization file or force all dates to be entered using four digits. To be Y2K-ready, Oracle's management console Enterprise Manager must be Version 1.4 or later, and Oracle Lite must be Version 2.5 or later. (Versions 2.0 through 2.3 shipped in 1996, so there are sure to be some of those still around.)
Versions of Oracle Express that are 5.0 or later are Y2K-compliant. Microsoft SQL Server Microsoft identifies a number of minor operational problems with its SQL Server database.
Leap day is again a problem: SQL Server's Enterprise Manager didn't allow us to schedule jobs on Feb. 29, 2000--the day didn't exist in its date spin control.
In addition, database backups will not perform a date check correctly after Jan. 1, 2000, and will overwrite old backup media without the normal "Do you really want to do this?" warning.
According to Microsoft officials, these problems will be addressed with SQL Server Service Pack 5, which is expected to be out before the end of the year. SQL Server automatically adjusts two-digit years when they are entered to handle the change to the new millennium: Any figure less than 50 is treated as being in the 21st century; any figure greater than 50 is considered part of this century. Sybase Adaptive Server Enterprise and SQL Server
Adaptive Server Enterprise 11.5.1 is the first fully Y2K-compliant release of that product.
Version 11.5 has most of Sybase's year 2000 fixes but still has a few bugs that prevent log analysis and session playback for dates in the year 2000 and later. Sybase SQL Server, the predecessor to Adaptive Server Enterprise, must be at Release 11.0.3.2 to be Y2K-certified. This release included fixes for tape backup problems similar to those that also affect Microsoft SQL Server.
As of its year 2000 releases, both Sybase databases interpret two-digit dates of less than 50 as being dates in the 21st century. The workgroup database SQL Anywhere needs to be at Version 5.5.02 (which was released in mid-1997) or later to be year 2000-compliant.
IBM DB2 DB2 2.1 and later versions for Unix and Windows NT (including DB2 Universal Database 5.x) are certified by IBM as Y2K-ready. DB2 2.1 shipped in 1996. On the mainframe, DB2 for MVS/ESA needs to be at Version 3 or later (with some service upgrades). Version 5 contains all the required fixes already.
In the data warehousing space, Visual Warehouse 2.1 or later is also Y2K-ready. For more information on DB2 for the AS/400 and DB2 on MVS or OS/390, visit IBM's product software readiness database at wwwyr2k.raleigh.ibm.com. (We frequently found it too busy to service our requests, however.)
Sifting through the options for dealing with unstructured data
Content management/ collaboration tools |
Advanced search tools |
Data warehouse tools
In search of knowledge
Sifting through the options for dealing with unstructured data
Given vendor allegiance, DMA could bring order to document management
Products from different categories aim to solve the problem. Each is focused on the capture, storage, retrieval and analysis of documents, text, images, video and forms of other unstructured data. Here are some examples: Content management/collaboration tools
Verano Makes Iluminar, a document management system that includes parametric and context-sensitive searching of contents. Iluminar works by extracting metadata from documents and other types of unstructured data, then allowing for searches based on author, project name, expiration date or other characteristic.
Enigma In October, began shipping Xtend, which builds on RTF, SGML and XML tags embedded in reference documents to allow users to conduct topic-specific searches and create hyperlinks between purchased reference material and internally generated documents.
Cypress Built around a universal printer driver that can take input from files and documents stored in different page description languages and on a number of platforms, including mainframes, minicomputers and PCs. Cypress extracts metadata about the documents and files and stores it in a repository that can be searched by Knowledge Builder, another component of the suite.
IntraNet Solutions A Java server-based document management system that allows users to view diverse document types in HTML, PDF or proprietary formats. Also includes a component, called Legacy, that scans and converts paper documents to Web-ready files and maintains them in a repository. Inte-grated with the Verity search engine.
Documentum Makes Enterprise Document Manage-ment System 98, a Web-accessible document and image management and workflow system that includes integration with enterprise applications such as SAP AG's R/3 and PeopleSoft's Manufacturing module.
SageMaker Targeted initially at petroleum, energy and utilities companies, SageServer Windows NT-based repository and SageWave browser allow for aggregation of text, graphics, maps, charts and other unstructured data from external and internal sources. SageServer is built around industry-specific data models that impose common meanings on data in documents and other unstructured content. Data is tagged and accessed using XML. SageMaker plans to expand its product focus to financial services and other industries.
Advanced search tools Verity Recently enhanced its search engine with the Verity Profiler Kit, which allows individual users to scan large document databases using business rules rather than keywords.
Virage The company's Video Cataloger derives searchable metadata from colors and shapes in a video signal. Users can then search video content for specific images or other information. Video Cataloger is like Adobe Systems' Photoshop for video. Virage also recently announced AV Photo Finder, for searching digital photographic and other images. This was developed with Compaq Computer's
PC Docs/Fulcrum Combines the PC Docs document management system from parent company PC Docs Group International with a Windows NT-based indexing, search and retrieval engine. Widely used by law firms and government agencies.
Excalibur Technologies Makes RetrievalWare, a Web-accessible search engine that can access more than 200 document formats. Recently, Excalibur enhanced the product, in Version 6.6, to access and retrieve information from e-mail and groupware products, including Microsoft's Exchange and Lotus Development's Notes. Also sells Screening Room for searching video.
Relevance Technologies Recently acquired by Documentum, Relevance's Semantic Modeling Architecture allows for indexing of a wide range of text-based content, including documents generated by ERP and sales force automation systems.
Data warehouse tools NCR Has announced plans to extend its Teradata large-scale data warehouse platform with an object repository capable of storing and providing access to documents and other unstructured data. The object repository, which is currently in development, will be applied first to customer relationship management applications that will allow customer support representatives, for example, to quickly access documents with customer information as well as transaction records.
Note: List is representative, not all-inclusive
