Follow Datanami:
August 14, 2012

What Real-Time Really Means

The term ‘real time’ gets thrown around a lot when discussing analytics. While it is okay for big data platforms to take their time when providing long term strategy solutions, businesses need quicker responses when it comes to fraud detection, outages on a smart grid, et cetera. Vendors are quick to offer ‘real time’ results but too frequently the enterprise’s definition of real time differs from that f the vendor.

To better lock down that definition and more, Philip Russom, Research Director of TDWI’s Data Management Department, and Dale Skeen, CTO and Co-Founder of Vitria Technology, delivered a TDWI sponsored webinar on Tuesday. Discussed were the differences between operational intelligence (OI) and business intelligence (BI) as well as examples of existing real-world OI applications built to run in real time.

Russom noted the differences between OI and BI. According to Russom, OI deals with “true real time,” which he discussed later, while BI was rarely even “near time.” Put simply, OI is significantly faster. However, he was by no means denigrating BI. Rather, he recognized that BI and OI were separate tools for separate tasks. While OI is faster in the tasks it is meant to be, it lacks the scalability required to produce long-term business plans that BI can.

Of course, it is important to define in concrete what it means to deliver in real time. According to Russom, real time responses reflect not only on absolute time but also on the various types of data. For Russom, in order for a response to be considered real time, it has to combine the event data with existing data and process it within milliseconds. He advises that queries that need to be answered within fifteen minutes should be answered with OI.

Skeen delved further into the specifics of OI and the advancement of real time, highlighting an interesting perspective on just how much quicker transactions have happened in the last ten years. For example, according to Skeen, in the last decade trading analytics went from happening in 30 minutes to a tenth of a second. Similarly, document transfers have decreased from three days to 45 seconds, data warehouse refreshing from a month to an hour, and phone activation from three days to thirty minutes.

Skeen put this all in context via a specific use case of operational intelligence in action when he discusses telecommunications grids. Skeen notes two real problems, “Maximizing customer experience for VIP customers” and “Predicting cellular equipment failures,” which OI have helped solve. The solution, which is referred to as “Service Assurance for Mobile Telco”, handles 250,000 mobile events per second (most of them irrelevant) and processes them through a CRM before they are delivered to a virtual dashboard.

The virtual dashboard shows spots of higher than anticipated cell usage before said usage causes a shutdown. These projected shutdowns are caught by Hadoop HDFS. According to Skeen, while Hadoop is generally considered more of a BI structure, it is remarkably good at catching systems failures before they happen.

Per Rossum and Skeen, operational intelligence does not merely outclass business intelligence in speed. It is rather a different tool set for a different problem set. While both note that OI is increasingly scalable and BI is increasingly fast, they have their different, complementary functions. Until they meet up and accomplish the same goals, that is they will remain.

Related Stories

Using an In-Memory Data Grid for Near Real-Time Data Analysis

What it Takes to Deliver Real-Time Traffic Info


Tallying Hadoop’s Hard Numbers