A peek into the tech behind Amazon, Flipkart annual sales

E-commerce giants are banking on India’s festival frenzy, which always attracts massive online traffic and sets the cash registers ringing. Flipkart’s Big Billion Day Sale, Amazon’s Great Indian Festival Sale, and the new entrant PayTM are battling for the biggest pie in India’s online festive sales.

This is certainly not the time when any online retailer would want to deal with server outages, which pushes its customers towards competition. Yet, it happens.

After Flipkart’s servers crashed during the Big Billion Sale in 2014, the company had to apologize to its customers. This year, reports have emerged of Amazon India website suffering outage due to massive surge in traffic.

According to analyst firm Forrester, despite some hiccups in 2016, India remains the fastest-growing online market in the world. The value of online sales in India will hit the USD 48 billion mark by 2020. Forrester had earlier predicted the value to reach USD 75 billion; but had to slash it post demonetization.

So it should come as no surprise when Amazon founder Jeff Bezos committed to invest USD 5 billion in India, and Flipkart saw a slew of investors with the latest USD 2.5 million boost from Softbank. 

Too much pressure for the IT department to keep the show running? You bet. CIO India spoke to industry experts to find out how the tech departments of online retail companies prep for big sales and what happens when disaster strikes.

K. Vaitheeswaran, author of Failing to Succeed: The Story of India’s First E-Commerce Company, said, “The only question that needs to be answered is about scale. Traffic and transactions will be there and grow significantly. Businesses have to estimate a number that determines how much their traffic will grow during these 3-4 days. And then prepare IT systems to manage the highest potential load for these days.”

If a crash happens, it means retailers didn’t get the initial estimate right, he added. In case of Amazon, AWS allows you to automatically scale up the number of instances based on the traffic – but even that has its limitations.

According to Vaitheeswaran, in the case of a crash, a key metric that helps online retailers come back up on their feet is the number of concurrent users. It means the maximum number of users expected to be on the site simultaneously during the sale. When the site crashes, this number becomes zero and that’s when the tech department gets into action to put the system back up quickly.

 “In most most of these cases, they can just restart the system to fix the issue or they can use additional resources if there is a sudden unexpected surge in traffic, he said.

“The tech team creates additional instances, which could be more server space deployed for the sale. So, when the system comes back again, they have already created new server space and they are able to manage the surge without experiencing downtime.”

Sriram Manoharan, Founder and MD at Contus, highlighted the importance of testing the infrastructure before D-day. “Testing is very critical. You can’t wait for the event to happen or for it to breakdown,” he said.   

So how does IT manage the massive traffic surge while ensuring a successful sales day? “These days, technology such as microservices are being used to deal with this. Earlier, whole applications would run on single instances; but now we use containers and divide the whole application into multiple components, which helps in avoiding 100 percent downtime,” Manoharan said. 

But it’s not just IT that needs to run without glitches for a successful sale.

Sandy Shen, Research Director at Gartner, points out how supporting systems need to stay in sync with IT. “If the application is bought from a vendor as a SaaS, then the vendor needs to ensure their SaaS offering can support the volume and velocity of transactions. If it is an on-premise deployment where the business manages everything themselves, they need to ensure their datacenters have the capacity to handle large transactions, and the supporting systems e.g. marketing, supply chain and logistics can all work in a seamless fashion,” said Sandy Shen.

So, now you have arrived at a number that tells you how much traffic your site will experience, you have tested the infrastructure, and you have compartmentalized applications to ensure nothing goes wrong. This is when the festival frenzy comes into the picture.

According to ASSOCHAM, India’s internet population is constantly growing and expected to reach 600 million by 2020. You had planned for x number of users, and now you get hit with 50x. How does IT handle this and what should be the contingency plan in case of an outage?

“A drop in performance or system having down time would have a huge adverse effect on the perception of shoppers,” explained Finny Chellakumar, Practice Head – Ecommerce, Aspire Systems. They key is to always have a communication plan in place.

“While the backup is getting in place, you can engage your users through means like Gamification, Counter etc,” he added.    

Another trick is to play down your traffic and divert it from your site to an associated site while the main site goes down. “This way the number of hits your site would get could be reduced while you get to work on it,” said Chellakumar of Aspire Systems.  

Behind the checkout clicks and online carts full of festive offers, there is an equally frenzied set of faces working to make the experience worthwhile to customers. Because well, the show must go on.