With the inevitable zombie apocalypse, having the right strategies to combat the plague will be essential. Turns out that trouble-shooting application performance isn’t much different. As any good zombie fighter will tell you, in a pandemic that threatens to consume all humanity, it’ll be important to find the first person infected – called “patient zero”. Knowing that sucker's history can help determine how and when the infection started, and with a bit of luck, a way to stop it. You might scoff, but there are many parallels between this and the way we manage application performance. Ok, perhaps not on a World War Z scale, but still troublesome enough to bite your business where it hurts most ...
The 2010 FIFA World Cup fever tested the Internet’s limits more than ever before. News site traffic reached a blistering 12.1 million visitors per minute – a record that far exceeds the 8.5 million set by Barack Obama’s presidential election win back in 2008.
And, this year, the Internet is taking it one step further as the BBC plans to host a 24/7 World Cup feed, across all television, radio and digital platforms. That’s 50 percent more coverage than 2010. So, with more than 160 hours of programming, including highlights and match replays across all of their online channels, you have to wonder, how are they going to pull it off?
DevOps will be conducting some pretty rigorous testing to ensure their channels can hold up under what could be another record-breaking moment of traffic in Internet history. But, will this be enough?
A key to performance testing is being able to simulate peak traffic to ensure your website will hold up under load. But, it’s important to avoid the all-too-common mistake of only testing within your corporate local area network (LAN).
Viewers of this year’s World Cup will span continents, so only testing traffic capacity within your own network will not suffice. It’s great if your site is able to sustain one million concurrent connections on your LAN, but when those connections are coming from other regions, putting more strain on your bandwidth, performance becomes uncertain.
Simulating a load scenario where the traffic only originates from within the corporate LAN can be compared to training for the Tour de France … on a stationary bike. Sure, you may be able to tackle the 3,500 kilometers over 23 days of training, but that doesn’t account for friction on the road, cyclist traffic or natural elements like wind, heat and rain.
That kind of training is only testing your body’s ability to perform under the most ideal conditions, which is the same as testing website performance from within the corporate LAN. On the LAN, you don’t have to go through the firewall, cache, load balancer, network equipment, modem or routers, thereby avoiding any kind of packet collisions or re-transmits. Ideal? Yes. Realistic? Not a chance.
Cloud-Based Performance Testing
Cloud-based performance testing enables broadcasters to simulate the millions of real users coming directly from the Internet – just as they will be on June 12 when the World Cup kicks off.
The cloud is extremely well-suited to generating the peak demands required for website performance testing. Not only can you ensure that sufficient compute power is available to scale from 100,000 to 1,000,000 virtual users and beyond, but you can also do it on demand with automatic resource provisioning.
Gone are the performance-testing delays of deploying and verifying internally managed hardware. With the cloud, concerns over the number of available servers on hand and whether idle servers are wasting valuable resources are something of the past. Performance testing can be run from anywhere with an Internet connection and a browser without the risk of costly over provisioning.
If broadcasters like ESPN, the BBC and ITV that are expecting to handle an increase in traffic from the World Cup were to solely use an on-premise testing model, they would have to acquire enough resources to support the tremendous capacity planning for that event. But, those resources could potentially go unused for the rest of the year.
Matters are complicated further when you consider that viewers will expect to watch seamless coverage of the games on TV, tablets and smartphones, so traffic simulations should take multiple devices into account.
The elasticity and agility of cloud resources means they can be easily scaled up or down as needed while only paying for what you use thanks to pay-as-you-go or utility-style pricing. This makes it an extremely efficient and cost-effective solution for performance testing needs.
Handling Global Load
Performance tests for something as big as the World Cup need to go even further to test global demand from most countries around the world. After all, soccer is one of the most widely watched sports there is, with a footballer fan base extending far beyond this year’s host country, Brazil. The global nature of the cloud serves this requirement well. Load tests can easily be carried out across different geographies since the cloud allows virtual users to be replicated in a variety of locations to test international performance. Cloud providers and test solutions can evaluate website global readiness, all without requiring you to stand up an expensive data center of your own in each location.
All in all, it would appear that technology is saving the day once more. The ability to broadcast live international coverage over the Internet enables an increasing number of fans to get connected and stay connected. With that, broadcasters let themselves in to a bottomless pit of demand for live viewing - which, in turn, leads to increased revenue from advertisers. Without cloud-based performance simulations, chances are, broadcasters would be getting yellow cards of dissatisfaction all around.
Alon Girmonsky is CEO of BlazeMeter.