Better test your systems rather than test your customer's patience!

The Flipkart episode finally got me back into blogging mode. I am sure they must have tested their IT Systems before going into the big sale day. I wonder what tests were done to assess uptime and availability capabilities they carried  for this Mother Of All Events.

I remember way back in 2005 I went to my CEO stating the need to do a stress, endurance and resilience tests before we launched the online trading website. These tests are expensive. I remember the popular vendor in those days in India who had the capability on load runner, pricing services with license cost on the go plus a monopoly premium ! The sales head and the operations head mocked me in the MANCOM saying that IT builds systems that they have little confidence in, and need additional money to validate their creation!  Partly, I agree with them; as in those days; we used to underestimate and under budget the testing processes. Testing was limited to functionality testing and was done by internal users. Availability, uptime and scalability was just a infrastructure sizing problem! 

For a couple of years in the later half of the first decade of 21st century, oversizing infrastructure covered up coding in-efficiencies and database design inefficiencies. With the passage of time, IT teams have started testing application code for scalability issues apart from just security loopholes. I was quiet amazed that in this day and age, Flipkart comes up as a total goofball on this account! 

1) Firstly I doubt if any serious testing was ever done on availability, resilience and endurance. It takes about seven weeks of three iterations and analysis to get anything right , even with the most advanced tools and analysis of results. Every bit of application code needs to be tested for behaviour under loads when you intend to take on a mammoth activity like FlipKart did.

2) Second point. Everyone knows that no one can provide for infinite resources at any given point of time. Sound economic activity like retailing business will not allow massive oversizing for a single day's event. But with IaaS providers, one could have made arrangements for infrastructure to be provisioned on demand...but again don't know if flipkart online shopping application is cloud ready. I guess every e-channel better start getting cloud ready.

3) Third point. Everyone knows that no one can provide infinite resources at any given point in time. If one were to place faith in their business model, and the fact that it will pay off, one could have always drawn a line and declared a finite number of users assured entry on a first come first serve basis and the rest could queue up for entry. One could have played a bit with the pricing model as well.

Point is availability of service in the online world is not an IT problem alone. Business will need to come up with sales strategies and pricing models and footfall management to work within the cost structures to maximise profits on these massive sales festivals.

I am however amazed at the lack of adequate preparedness by Flipkart before they put their brand on the chopping block!