InformationWeek
Defining The Business Value Of Technology
September 14, 1998
Microsoft And OLAP: Sudden Impact
continued...page 3 of 4
Microsoft's DTS is also geared to work with SQL Server, and it, too, does a serviceable job. The chart in the PDF file compares it to incumbent data extraction and transformation tools.
A user can set up a pretty good SQL Server 7 data warehouse management system by combining the abilities of SQL Server 7's enhanced Replication Services and Microsoft Repository's new storage of DTS packages and schedules. This will be reasonably, but not entirely, attractive to pure-Microsoft shops.
First, data extraction and cleansing are consistently underestimated for both cost and time on data warehousing and data mart projects. The Conference Board study reflects this: Only one-third of projects used automated extraction and cleansing tools. Second, many data warehousing systems cross divisional, and even organizational, boundaries. This forces IT to create interfaces between a number of other target data marts, data warehouses, and OLTP databases. Systems from ETi, Prism, and many others designed to connect heterogeneous data sources give developers much more flexibility in these situations. Finally, there are other features, such as run-time performance, breadth of interfaces offered, and utility features, that make other data extraction and transformation tools valuable still. Many tools add modeling or repository capabilities (see chart, p. 318) that simplify scripting or warehouse administrative routines.
In sum, the free SQL Server 7 data warehousing tools are well targeted and strongly supportive of working with that database, but they leave ample room for vendors of other data warehousing tools to add value to the warehouse design, building, and maintenance processes. It will be interesting to see if OLAP vendors will have the same room to maneuver in their market.
OLAP Servers
The size of the OLAP market is growing from $1.4 billion in 1997 to $1.8 billion in 1998, according to the OLAP Report. This is a 30% annual growth rate and equal in size to the data warehousing market. No wonder Microsoft wants in, but why and how long when it's giving away its main product? Now Seagate Software, with its recently released free copy of Seagate Worksheet, which has OLE/DB for OLAP capabilities, has turned the client-side desktop OLAP market topsy-turvy.
Nonetheless, the underlying situation is that OLAP deployment trade-offs (performance, function, and storage) currently drive the OLAP market into natural niches. No single technology dominates.
ROLAP uses relational databases for storage but special, denormalized star or snowflake designs to store the analytical data. ROLAP has the advantage of being able to store the huge amounts of data required in some market analysis systems. ROLAP can perform fairly rapid queries and simple summaries, but its downside is poor performance for the types of complex calculations associated with financial modeling, analysis, and forecasting.
Multidimensional OLAP (MOLAP) is based on storing data not in relational tables but in special cubes with precalculated summaries. MOLAP has become increasingly popular, as refinements in storage of multidimensional cubes have increased not only in maximum size but also in performance. Although MOLAP cannot match ROLAP for compactness of database, the relatively fast response time of MOLAP software (seconds rather than minutes for many operations) has won over many users. In addition, MOLAP linked to the right client software can deliver complex financial and other modeling calculations with consistently better response time than similar-sized ROLAP systems.
Hybrid OLAP (HOLAP), neither purely Relational nor Multidimensional, has appeared in the market over the past few years. HOLAP allows the size of the multidimensional database to be larger (ranging from three to 10 times, but still far from matching ROLAP) while generally preserving the response time. MS/SSOS has just such a design.
Recently, many vendors have added lots of new features to their desktop OLAP clients--which often link back to MOLAP, HOLAP, or ROLAP servers. Those features include simplified or even English-like queries, and different data views, such as pivot tables, briefing books, and 3-D graphics. Some also have the ability to partition or locally store models for standalone use, refinement, and later write-back to the main model. Writing back is important for planning and budgeting processes. In sum, the OLAP market has developed natural niches associated with distinctive product features, which meet special needs.
continued...page 4
return to page 1, 2
Back to Labs
Send Us Your Feedback
Top of the Page
Featured Microsite
