Blog: Mike Ferguson Subscribe to this blog's RSS feed!

Mike Ferguson

Welcome to my blog on the UK Business Intelligence Network. I hope to help you stay in touch with hot topics and reality on the ground in the UK and European business intelligence markets and to provide content, opinion and expertise on business intelligence (BI) and its related technologies. I also would relish it if you too can share your own valuable experiences. Let's hear what's going on in BI in the UK.

About the author >

Mike Ferguson is Managing Director of Intelligent Business Strategies Limited, a leading information technology analyst and consulting company. As lead analyst and consultant, he specializes in enterprise business intelligence, enterprise business integration, and enterprise portals. He can be contacted at +44 1625 520700 or via e-mail at mferguson@intelligentbusiness.biz.

June 2008 Archives

As we head down the road towards real-time event-driven computing and move gradually away from batch processing it is clear that the number of events in some vertical industries is going to be enormous. Not surprising therefore that if we are looking to analyse this data before it reaches a data warehouse then in-memory data is going to be essential to scale up complex event processing (CEP) and business activity monitoring (BAM). You could argue that as solid state disk starts to replace spinning disks that it will all be in memory in the not too distant future. In my opinion we are going to start to see parallel query processing and scoring happening in parallel on in memory data. A good example of this is the strategic partnership emerging between Teradata and SAS which offers deployment of scoring models developed in SAS in the Teradata DBMS. This kind of functionality is the beginnings of massively parallel scoring on in memory data. Also other database vendors such as IBM and Oracle are investing in in-memory extensions to their database products.

If you are looking at scaling CEP or BAM on large volumes of events then it would seem that in-support for handling memory data is going to be high on your shopping list


Posted June 24, 2008 7:16 PM
Permalink | No Comments |

For those of you looking at Operational BI a very common question is how will this use of intelligence impact the classic BI system architecture? To answer this question requires that we first define operational BI. This is the use of intelligence in every-day business operations. We are not talking about running reports that access an operational system here. What we are talking about is much more comprehensive than that. Two major used of BI in operations are:
• Access to on-demand BI in the context of a process activity
• Event-driven automatic analysis and decision making

The on-demand use of BI in business operations is about right-time intelligence. This is where a user chooses to perform a specific activity as part of an every-day operational business process and is presented with only the BI that is relevant to help the user perform the task more effectively. It is sometimes referred to as a ‘BI in context’ and is the most precise use of BI. To make that happen typically requires the deployment of BI systems in a service oriented architecture (SOA). In that sense it requires companies to bring together IT professionals responsible for business process management and SOA with IT professionals responsible for BI. Most companies are not yet organised in this way and so will need to mobilise to make this happen. However the impact on the BI system is relatively minimal in that the latest releases of many modern BI platforms today are already ‘service enabled’ i.e. they have web service interfaces to allow on-demand invocation of reports and queries from composite applications, processes and portals. So in this case the BI system per se is really being extended to plug into a SOA and the emphasis is more on re-organising IT to make on-demand BI happen.

However event-driven automatic analysis is a different ball game. In this case we really are stepping outside the ‘classic’ BI system architecture in the sense that this aspect of operational BI is about being able to detect operational events, automatically analyse data and take action well BEFORE the data reaches a BI system. This brings together new technologies such as data streaming, highly parallel in-memory data and rules engines with familiar technologies such as event driven data integration and scoring models built by power users using data mining tools. So if you have event-driven data integration tools already (most ETL tools support this capability) and you already have power users developing predictive models using data mining tools then there is nothing new in these areas. However it is the data streaming, in-memory database (both IBM and Oracle for example have extended their DBMS products by integrating them with in-memory databases) and rules engines that are new. Implementing operational BI in this way is about automating business optimization in every day operations. This kind of automation is operational performance management. So some BI system related technologies are used in this new form of operational performance management while other required technologies are brand new to us. Event-driven business monitoring therefore is a different architecture to a classic BI system because of the need to analyze stream data in memory before it ever reaches a data warehouse. Once analysis has taken place, decisions made and action has been taken it is only then than that event data may well find its way into a BI system via classic data integration processing. So we have to think differently about this. In that sense it would not surprise me if this form of real-time business event monitoring was initiated as a new project with a different architecture outside of a BI team.

In many cases it may well be that this kind of sponsored project is part of a business process management or an RFID initiative rather than being part of a BI program. If this happens it should in my opinion be quickly flagged up and brought under the management of the IT team responsible for BI and Performance Management otherwise the worlds of operational performance management and strategic performance management will never meet. This would be a very disappointing outcome. Frankly, they have to meet because business users need performance management to encompass both the classic strategic performance management (strategy management, scorecards, dashboards, budgeting, planning etc.) AND operational performance management. They need to see what is happening over time in their area of responsibility alongside what is happening right now. Therefore once again we need to keep our eye on the ball regarding project sponsorship, ownership and architecture so that the true benefits of on-demand and event driven operational BI becomes a major contributor to business performance and allows people to leverage intelligence to help them optimize every-day business operations.


Posted June 18, 2008 12:39 PM
Permalink | No Comments |