Semantic Metadata in Reporting Tools

You can contact the author, Barney Finucane, by e-mail if you have any comments, observations or user experiences to add. Last updated on December 18th, 2009.

Introduction
Storing and administering the model
	BusinessObjects Universes
	MicroStrategy
	Cognos
	Oracle BI Server
Structure
	BusinessObjects Universe
	MicroStrategy
	Cognos
	Oracle BI Server structure
Data federation
	BusinessObjects
	MicroStrategy
	Cognos
	Oracle BI Server
Performance
	BusinessObjects
	MicroStrategy
	Cognos
	Oracle
Caching
	BusinessObjects
	MicroStrategy
	Cognos
	Oracle
Access via external tools
Conclusions

Introduction

In 1996, Business Objects was awarded a patent for the idea of a ‘semantic layer’, based on its universe technology. The universes provide a dimensional model of the data source the system accesses and are designed to shield end users from the complexity of corporate databases. At the same time they provide these users with access to the data in them and maintain role-based security. Another great advantage of semantic layers is that they provide consistent names for data objects. They also provide more consistent data, since all the users are accessing the same data source.

Since Business Objects patented the idea, several vendors have adopted very similar ones – notably MicroStrategy, Cognos and nQuire, which came to Oracle via the Siebel acquisition. Both MicroStrategy and Cognos were involved in patent infringement law suits with Business Objects because of this. So was Brio, which settled with Business Objects for $10m, before being acquired by Hyperion. As a result, Hyperion acquired rights to the semantic layer patents, and was developing its own version when Oracle acquired it. Oracle therefore owns those rights, as does IBM, through its acquisition of Cognos, which had paid Business Objects $24m to settle.

The products differ in detail, but their overall structure is much the same. They offer a dimensional model on top of a physical layer. In fact the original patent specifies that the model contain at least one dimension. Semantic layers treat measures as a separate object type from elements of other dimensions. The measures are either taken directly from the data source or calculated.

These products are not databases. The main role of the semantic layer is to provide a model which is used to define a query. The products do provide various degrees of caching, but this is not a central element of the design. The semantic layer requires modelling by a user with strong IT skills. The original intent of the semantic layer was to free business users from IT, but this has not succeeded. Indeed, the products described here all require a significant amount of IT support to operate.

There has been a good deal of discussion over the years on the compatibility of metadata information from competing vendors, but the products do not offer it now and there is little sign that they will in the future. It simply is not in the interest of the individual vendors to open up their systems to other, potentially predatory vendors. Furthermore, small logical differences between the products make it difficult to provide automatic translations. Master Data Management (MDM) products also make some attempt to solve this problem, but coverage is far from complete, and the issue of metadata format is only one of many.

The purpose of this document is to provide a brief overview of the most popular products that use the semantic layer concept. They are among the most widely used reporting tools on the market, which is a good indication of how successful the idea has been.

We do not address Microsoft, although Microsoft arguably has two candidate products for this discussion. The first is Reporting Services, which comes with its own simple semantic metadata features. In the 2005 release it looked as if Microsoft was intending to expand this semantic layer significantly. At the time of writing Microsoft seems to have changed its mind, and no longer emphasizes the semantic layer of Reporting Services and they are too weak to be considered in detail here.

Microsoft Analysis Services also offers ROLAP base on its cube models. Analysis Services can also be viewed as a semantic layer, data cache, aggregator and multidimensional calculator. You could build an Analysis Services application for little more effort than it would take to build and tune these relational semantic layers, and then get much faster performance and richer analytical capabilities. But we regard Analysis Services as a separate class of application and treat it in another document.

Subscribers can click here to read on.

To view the remainder of this article, you will need to purchase a subscription to The BI Verdict.

Subscribers gain access to:

Product Evaluations - Detailed and in-depth evaluations of dozens of leading business intelligence platforms and tools.
Market Analyses & Commentaries - A series of in-depth analyses and insightful commentaries from a vendor-neutral perspective.
The BI Survey summary - Key findings and product dashboards based on the results of the world's largest independent survey of BI end-users.

Click here for more information about The BI Verdict.

Click here to view the full list of content available to BI Verdict subscribers.

Semantic Metadata in Reporting Tools

Contents

Introduction