The Show Must Go On: Ticketfly Deals with Massive & Instantaneous Demand Spikes for Event Ticketing
Ticketfly is a progressive ticketing company that makes it easy to discover events, buy tickets, and share events with your friends – all with world-class technology and customer support. Ticketfly works with some of the best venues and promoters in North America. Ticketfly posts three digit revenue growth, year over year. As it gears up for another summer concert season, it can expect its systems to undergo immense stress, requiring a 10X increase in site performance. In addition, ticket ordering is a spiky business. Some events attract demand that is literally instantaneous. The resulting performance curve isn’t a curve but rather a huge spike, almost like what a denial of service attack would look like. This is how the business is: light loads and sudden immense loads, with users constantly hitting the refresh button to see if they managed to secure a ticket to an event that has a lot of demand and just began selling. To make sure Ticketfly works well, the company decided to simulate these spikes in traffic, measuring and tuning the underlying systems as a result so they can successfully serve customers in the next traffic spike.
Ticketfly hosts its own datacenter and its system is fine tuned to support user demand.It also has an overflow mechanism to Amazon Web Services. Tickets buyers come through a proxy, hosted on Amazon ec2. dubbed the lounge or waiting room, till they can access the cloud based system and make the purchase. The waiting room scales on demand. Site performance isn’t a nice to have feature, it is a must have, directly impacting customer satisfaction, since event attendees have zero tolerance for slow or unresponsive systems. Even though Ticketfly is one of the fastest online ticket sellers (in terms of time it takes to complete the purchase) on the Internet, users still hit refresh and really burden the system at times of peak loads.
As Ticketfly rapidly grows its business, performance testing is becoming ever more important. Ticketfly’s original testing set up was built on Amazon Web Services and used JMeter scripts along with New Relic. Switching over from Amazon to BlazeMeter was easy. BlazeMeter provided immediate scale to build up the features that typical performance testing requires. The switch meant that Ticketfly didn’t have to configure testing platforms anymore, making the move cost effective. It could also re-use its already existing JMeter scripts.
Ticketfly uses BlazeMeter together with New Relic. "The New Relic integration is fantastic because you can see the load live, the number of concurrent users, the requests per second, on a time-series view" said Spencer Greene, VP Engineering, Ticketfly. "We want to look at a slice of time and see how our distributed system is performing as a whole. What’s getting stressed? database? app servers? the purchase flow? You don’t just use performance testing to discover your ceiling, but to see where we need to apply our engineering resources.”