In-Depth

Microsoft Convenes Second Annual BI Conference

Microsoft touts an in-memory, column-based Excel data store on every desktop

This week, Microsoft Corp. is hosting its second annual Business Intelligence (BI) conference in its own backyard -- Seattle.

The big news out of Microsoft BI 2.0 is also the most forward-looking: Redmond announced "Kilimanjaro," its code name for a still-incubating, BI-centric version of SQL Server that it plans to deliver sometime in the first half of 2010. (SQL Server Kilimanjaro isn't to be confused with "Kilimanjaro," the one-time code name of Microsoft CRM, which was subsequently rechristened "Titan.")

Paired with an in-memory, column-based Excel add-in -- code-named "Project Gemini" -- Kilimanjaro is Microsoft's latest stab at pervasive (in Redmond's parlance, "people-ready") business intelligence.

Microsoft also discussed Project Madison, its effort to ingest the assets of the former DATAllegro Corp. Microsoft is working to integrate DATAllegro's massively parallel processing (MPP) capability with its SQL Server 2008 database. Not surprisingly, officials confirm, Microsoft doesn't plan to resell branded data warehousing (DW) hardware; instead -- and taking a page out of competitor Oracle Corp.'s book -- Redmond will market DW "Reference Architectures" in tandem with vendor partners -- including Dell Computer Corp., Hewlett-Packard Co., Unisys Corp., and Bull.

Industry watchers seem intrigued by the direction Microsoft plans to take its SQL Server-centric BI strategy. Once it ships, SQL Server Kilimanjaro could comprise Microsoft's most successful attempt to date to cater to Excel-centric end users while simultaneously combating the "spreadmarting" effect that is seen as part and parcel of Excel use in the enterprise.

At two years out, Microsoft's notional SQL Server Kilimanjaro and Project Gemini releases must still be regarded with skepticism. However, neither appears to address any of several long-simmering questions about Redmond's BI strategy, such as its lack of an over-arching metadata plan.

As one industry veteran (who spoke on conditional of anonymity) puts it: "[T]hey have no infrastructure story around dealing with the data issues that arise, much like they don't today. It's the basic blind spot of Microsoft. [They don't have any] clue about how businesses operate, as reflected in business intelligence and data integration tools with no metadata."

Self-Service SQL Server

Two months ago, Microsoft unveiled SQL Server 2008 -- less than three years after it shipped its oft-delayed SQL Server 2005 data store. Redmond isn't resting on its laurels, however. Officials bill SQL Server Kilimanjaro, slated to ship in 2010, as Microsoft's most BI-feature-rich SQL Server release to date.

"The big announcement is on a release that's focused specifically on business intelligence, SQL Server Kilimanjaro," says Herain Oberoi, group product manager for SQL Server with Microsoft. "The broader conference theme is about thinking about business intelligence, so what we're calling 'people-ready' BI. When we think about our long-term vision, it's pervasive business intelligence, and what we mean by that is getting pretty much everyone in an organization talking about business intelligence and using business intelligence."

For example, he says, Kilimanjaro will focus on self-service amenities -- particularly for reporting. "We shipped our first self-service reporting tool [i.e., Report Builder] back in SQL Server 2005, and we added Report Builder 2.0 later [i.e., in SQL Server 2008]. In Kilimanjaro, we're going to take that much further, and focus on taking [Report Builder] 2.0 and adding new capabilities to it so that it's really an end-user tool in … [that] it's easy to use and it's intuitive, but also in terms of … people [can get] up and running with it quickly."

Oberoi calls this "grab-and-go reporting," a more alliterative take on the concept of self service. "Instead of having to build their own charts and data, [end users] could just search for a repository. They could just drag and drop that chart on to their report. If I'm the end user, I don't have to worry about binding the data set -- it's all done for me," he indicates.

An In-Memory, Column-based Store on Every Desktop

Another key to self-service is giving the people what they want. For most users, this amounts to BI which isn't obviously BI: i.e., canned reporting or analytic functionality that's exposed via a portal. For other users, it's a more auditable, traceable, and governable (but still inescapably standalone) Excel experience.

In either case, Excel-based analytics are at the forefront of Microsoft's Kilimanjaro push. It starts with the experienced Excel users who will be able to create and save analytic applications as Excel spreadsheets. From there, these applications can be consumed and managed by Microsoft's Office SharePoint Server (MOSS).

"You have a neat business intelligence tool called Excel that a lot of end users love. As a result, you have these mini Excel applications all over the enterprise. From an IT perspective, [this leads to] your typical spreadmart problem. What Kilimanjaro will do is focus on this problem of end-user empowerment but in a way that's manageable by IT," Oberoi says.

The solution, he claims, comes in the form of Project Gemini, a still-gestating initiative that will comprise an Excel add-on (in the form of an in-memory, column-based facility) and beefed-up integration with Microsoft's SharePoint portal environment.

"This [Project Gemini] has two particular pieces which really change the way that end users interact with tools like Excel and how they build business intelligence solutions," Oberoi says. "[T]here's an in-memory, column-based store. This allows an end user to effectively work with large volumes of data which they would normally have to connect to a server to do. Now they don't have to do that; the way that manifests itself is through Excel."

Microsoft's pitch, which amounts to putting a full-fledged column-based data store on the end-user desktop, isn't all that far-fetched. Last month, BI start-up Lyzasoft unveiled Lyza, a fat-client take on the same proposition (see (http://www.tdwi.org/News/display.aspx?id=9129). It's an idea that flies in the face of the thin-is-in Zeitgeist, but Microsoft (and Lyzasoft, for that matter) say there's a definite appetite for fat-client desktop analytic solutions, which (they argue) can help make users less dependent on IT.

"This gives [users] the ability to manipulate just a huge amount of data [in Excel]. Whether it's sorting, filtering, or creating links between different columns, what's really happening is that under the covers [a user is] actually building out a business intelligence model, but they don't know that that's what they're doing," Oberoi explains. "This is sort of a mini-business intelligence solution [that] allows them to do not just PivotTables but build this sort of interactive BI application. What you're saving is a model, an actual OLAP model, and you're [also] saving a presentation [layer] which you can access through the Web."

Enter the second of Microsoft's Project Gemini initiatives: SharePoint integration. "Because we're delivering this integration with SharePoint, the end user can now take this application and save it or share it out to a standard SharePoint workspace. So this takes this business intelligence solution that they've built and directly hosts it on a server. From there, if you're IT, you can monitor it; you can provision it; you can decide who gets access to it," he explains.

Missing Metadata

Neither Kilimanjaro nor Project Gemini do much (if anything) to address Redmond's lack of a coherent metadata story -- at least one that encompasses both SQL Server and its ever-expanding BI ecosystem. (The latter now includes data quality -- thanks to Microsoft's acquisition of the former Zoomix -- and master data management, too). Oberoi, for his part, demurs when pressed on this issue.

"The metadata story spans a couple of different uses today. [We have] metadata for the application itself, metadata for the data itself, metadata for the presentation layer," he says. "The broader metadata story across Microsoft, which also includes some of the work we're doing with the master data management team, that's something we're working on, but that's not going to be a focus for the [BI] conference."

Asked about using SQL Server Integration Services (SSIS) -- which is today able to extract metadata from source systems and enrich it as part of an ETL process -- as a centerpiece of Microsoft's metadata management strategy, Oberoi again demurs: "That's not something we're planning to talk about" during the conference.

Data warehouse architect Mark Madsen, a principal with consultancy Third Nature, says it's an issue that Microsoft must come to grips with at some point.

"[Their] stock answer is always that they have metadata in SSIS -- it's in XML files [that] you can access. [Getting at] BI [metadata is] a little harder, but [it's] there. They don't tell you about how you need a third-party product to make any real use of it," he points out.

With its expansion into data quality and MDM, Madsen concedes, Microsoft is at least starting to grapple with its metadata heterogeneity. At the same time, he argues, Redmond's boutique organizational structure could compound its difficulties.

"MDM and DQ are one area [inside of Microsoft]. DI [data integration] is another, and BI is actually in the Office group," he concludes. "They're finalizing features for releases 5-to-7 years out in that line of business. Imagine what that means for BI."

Must Read Articles