Fortnite, AWS, and the Importance of Monitoring

The Battle Royale game Fortnite has become a sensation amongst online gamers in no time at all. To explain it in simple terms, 100 players are simultaneously dropped into a battleground measuring several (in-game) square kilometers, and must proceed alone or as part of a team towards a random central point on the map whilst avoiding or confronting the other players. The last man or team standing takes the top spot and wins the game. It all adds up to an intense and at times hilarious experience that can last around 1-20 minutes.

The growth in popularity of the game has been epic from a 60,000 players on launch last July to 3,200,000 players in under nine months, and suddenly keeping the game up-and-running was going to require some pretty serious infrastructure.

From day one Epic, the publisher behind Fortnite, has like so many other large businesses such as Airbnb, Unilver, and Netflix relied on Amazon Web Services (AWS) to keep it online.
AWS gives Epic the ability to cope when player numbers spike; the difference in infrastructure workload might be up to ten times difference between the peaks and troughs.

Epic also takes advantage of AWS’s “availability zones”. These 55 zones are designed to ensure web services don’t lag in any one zone. Where one zone fails another simply takes up the baton. Fortnite currently runs across 24 of these zones.

This isn’t to say that AWS and the use of availability zones are infallible. In February of this year Fortnite experienced multiple outages which even AWS’s availability zoning couldn’t prevent.

It’s also worth remembering that whilst many companies such as Epic rely on AWS for its reliability and stability it’s worth remembering that Amazon itself can still have problems.

Just last month on Amazon’s Prime Day the rush for bargains not only brought Amazon down but impacted AWS. Whilst the AWS service itself continued to operate normally, AWS customers were unable to login to their accounts.

More serious however was the four hour outage in AWS’ US-East-1 region in February this year which saw over half of the top 100 internet retailers impacted. Many websites saw the performance of their sites impacted severely (Disney’s store took over 1000% longer to load than normal), many other sites went down completely; the same availability zone having similar issues again in May.

All of this highlights that even if you’re using cloud service providers such as AWS or Google Cloud that monitoring your website is as important as ever.