Microsoft Opens Up on Shiloh

Microsoft Corp. is beginning to talk about the details of Shiloh, the next version of SQL Server, due out in mid-2000.

Microsoft’s revelations are expected to add to the product’s expectations and to dash some of the negative rumors surrounding Shiloh. For example, despite published reports to the contrary, Barry Goffe, SQL Server product manager, says OLAP services have been and always will be part of SQL Server.

"OLAP today is integrated into the database. That will not change with Shiloh," Goffe says. "That's not something we've announced and we don't plan to [because] it will always be an integral part of the database."

There are a lot of new tricks that database managers can look forward to from next year's release of Shiloh. "One of the key things we're doing is taking advantage of some of the new features in Windows 2000," Goffe says.

The first is support for large amounts of memory. Using Windows NT Server 4.0 Enterprise Edition and SQL Server 7.0 Enterprise Edition, the database could access up to 3 GB of memory. With Shiloh on top of Windows 2000, a user will be able to access 64 GB.

Next, Windows 2000 will support four-node failover, compared with two-node in NT 4. You can have three machines running different jobs; if any crash, they will failover to the fourth machine. Shiloh supports this action so there is no drop in performance. Shiloh also will take advantage of Active Directory to make it easier for database administrators to manage hundreds or thousands of SQL Servers distributed throughout an organization.

Wayne Eckerson, senior consultant at Patricia Seybold Group (www.psgroup.com), says SQL Server is a good platform for applications such as data marts, but it still isn't ready to take on the enterprise. "Microsoft would say they can and they've laid down the foundation to do so, but it's still relatively unproven in that environment," Eckerson says. "It's getting there."

Aside from adopting Windows 2000 features, Shiloh will come with a few of its own bells and whistles, including support for Windows CE devices. Microsoft's Goffe says application developers are looking to push the envelope of palm devices, and they can now do this using the same programming languages they are familiar with to build health care, shipping, sales force automation or any line of business application that is useful in the field.

"We're putting a lot of effort into making sure the programming model is consistent across every version of the product, so in particular on the CE platform, the internal programming language is saying we support SQL queries," Goffe says. "Really, the idea is by sharing the same programming model across every version of the SQL model, [developers] can take more advantage of the CE platform than they were ever able to before."

Also integrated into Shiloh will be native XML support in the database engine. To get a look at how XML works with Shiloh, a sneak preview is available. XML for SQL Server 7.0 is a download that provides some basic functionality on top of SQL Server. Goffe says the idea is to give customers a chance to start testing out how some of this will work. The download can be found at the Microsoft Web site (http://msdn.microsoft.com/xml).

A big area of concern for Shiloh is knowledge management. Microsoft has invested in data mining for five years to put intelligence into the database. Part of this technology is available in Site Server 3.0 Commerce Edition.

This technology is supposed to help predict what customers will purchase next by determining past purchase patterns and comparing them with other customers’ patterns. This kind of solution is ingrained in the database, eliminating the need to configure third-party tools or to hire consultants to set the system up. Shiloh will deliver a data mining extension to OLE DB so developers can integrate data mining into an application to lie on top of the groundwork established by Microsoft.

The team at Redmond has worked on the data warehousing end of Shiloh by adding materialized views. In large data warehouses, this feature speeds up the repetitive reporting processes. Goffe explains: "If I have a receivables report that requires a fair amount of preprocessing to take place to run that report, I can create a materialized view that preprocesses the data and stores the results on a disk. So the advantage is now that I go to run my report, all the processing has been done for me. All the server has to do is read the results off a disk."

"It's essentially allowing you to put summary tables into a database ahead of time and manage them automatically," Eckerson explains. "The only problem is it takes longer to load the database because you have to create these materialized views."