IBM Analytics Appliance

The scheduled announcements today were about a new analytics appliance.

Steve Mills set the scene, pointing out that business optimization (a market in which IBM includes analytics projects) is getting client investment at 2x the general business automation market. Personally I think this is why decision management is going to be more and more important relative to business process management as it is more about optimization of business and less about automation. This kind of optimization work has been a focus for IBM for a while – in terms of R&D, consulting and software – but the visibility of it has really increased recently. IBM has made internal investments (System S – Infosphere streams) and purchases (ILOG and SPSS). Steve emphasized a $10B software investment and a 4,000 strong consulting force.

Today’s announcement is on workload optimized systems designed to deliver out of the box optimized systems – decrease the time to value for these kinds of complex systems. Scott Handy talked about the challenges of applying analytics and the need to simplify the underlying infrastructure and optimize that infrastructure for the workloads they see in customers. He identified three emerging trends that will be addressed, he says, by the announcement:

• Simplified, unified infrastructure (convergence of network, storage and a move to virtualized pooled resources)
• Workload-optimized systems
• Workload-optimized computing

Only the first is really a widely supported trend – the rest feels more like IBM’s internal focus. IBM works with clients to develop systems for customers optimized for different workloads and problem domains (they claim 8,000 such engagements). Based on this they have developed pre-integrated hardware, software and services optimized for a specific workload. This, of course, means optimized for a specific problem domain (like claims fraud). This sounds very much like the SPSS focus on decision management applications.

IBM sees three different workload types:

• Search and Query – requires workload spreading
• Predictive analytics – requires lots of memory and support for complex math
• Risk Analysis – optimized for specific models

Core announcements from Ambuj then:

• IBM Smart Analytics Systems
  New systems, optimized for analytics. “More” than an appliance because still allow incremental upgrades while still simplifying install, reducing the need for IT and speeding time to market. Each layer within the system has been optimized for a specific solution area rather than as a separate layer. In one example they claim a 3x performance boost out of the box and 50% less physical space for hardware.
• IBM Smart Analytics Optimizer
  Add on to optimize existing data systems.  In-memory system taking advantage of vector processing and new scanning and parallel processing technologies to minimize needs for indexing and administration. Q4 availability.
• IBM Customized Solutions
  Server/storage/software combined.

Going forward the acquisition of SPSS is going to be key, though they have nothing to say today.

Questions:

• Where’s ILOG in all this? Why is there no discussion of putting ILOG rules and optimzation on to these systems?
• How will IBM integrate the SPSS decision management applications into these systems?
• And if it does, will it use the SPSS rules capability or replace it with the ILOG one (as it should)?