Real time analytics take aways

2

My last blog post about the Analytics Camp 2010 conference relates to Farhad Islam's talk about "Real-Time Analytics, Getting results when you need it". Farhad proposed that changes must occur not only at the Business Intelligence vendor level, but also to both operating systems and hardware to increase response time of analytical requests. He stated that business intelligence vendors are already experts in addressing the data consolidation and analysis components, however they don't consider the user interaction component.

First, the main contributor to business intelligence system implementation success is NOT related to the vendors functionality but to the business intelligence implementers and administrators themselves. This dependence on individuals causes BI to be limited to human capabilities, including all the associated errors & delays.

Second, business intelligence vendors must take into account the utilization of reports and ad-hoc queries. If the vendor's product can complete analysis on the requests actually being made, it could then use these results to accomplish improvements in real-time response time. His suggestion is that the systems should dynamically and automatically analyze the frequency of similar calculations so that additional requests of the same nature are addressed immediately without waiting for the system to complete the same query again.

Also, Farhad reminded us that it is important to keep priority in consideration and use it to modify the query queue dynamically. Information on the person's level in the company, the work they are doing, the time of day, the processing length of the request, etc should all be used to ensure that the processing is working effectively toward the benefit of BI users. You wouldn't want the CEO of the company, who is on the phone with a customer, to have to wait even 1 minute to get a response.

Share

About Author

Angela Hall

Senior Technical Architect

Angela offers tips on using the SAS Business Intelligence solutions. She manages a team of SAS Fraud Framework implementers within the SAS Solutions On-Demand organization. Angela also has co-written two books, 'Building BI using SAS, Content Development Examples' & 'The 50 Keys to Learning SAS Stored Processes'.

2 Comments

  1. Some additional nitty-gritty details to the very well written article by Angela is provided below in case anyone is interested. --Farhad(1) My talk on Real Time Analytics proposed an end-to-end solution architecture for supporting real time response from Analytic servers. The request for analysis is assumed to originate from a remote "thin" client. A smart phone was discussed as an example of such a remote thin client.(2)Software agents were proposed to act proactively at the server and also at the smart phone client end.(3) The server end software agent would be responsible to change task priorities to support real time response.(4) The server end agent would interface between the heavy-duty Analytic software and the real-time OS to formulate a real-time response for the remote client.If anyone needs any further details, please feel free to get in touch with me: farhad919@gmail.com

  2. I thought that he brought up some really great points as well, and it is where I think analytics is heading, but we are not there yet. It is kind of a meta-analytics, where we use analytics to manage our analytics, but we need full control of the information eco-system to get it. We have a lot of work to do to get there.

Back to Top