How to play · 2 minutes

01Your app

Customers arrive in waves

You run an app, for example your shop. Customers use it all day. Sometimes a lot of them arrive at the same time. That is a wave.

The game counts customers per second. One customer is one request to your app, like a shopper opening the checkout.

Your job: serve every customer on time, without paying for servers you don't need.

Play it as
Customers arriving now0per second
Customers per second, over 5 minutesCustomers arriving

Quiet at first, then a rush, then calm again. Every team gets the same waves.

02Reading the chart

Purple is demand, black is what you can serve

The purple area is customers arriving. The black line is how many customers your instances can serve on time.

Black above purple: everyone is served. Purple above black: there is not enough room, and customers pay for it.

  • Amber: kept waiting. A slow answer. It costs $0.25 per customer.
  • Red: lost. An error, or no answer after 5 seconds. It costs $1 per customer.
  • Dashed and dotted lines: your max and min instances.
Customers per secondCustomers arrivingWhat your instances can serveKept waitingLostMax instancesMin instances
Money lost$0
Lost checkouts0
Cloud cost$0.00
03Instances

More customers need more instances

Your app runs on instances: copies of the app on servers in the cloud. One instance serves about 48 customers per second on time.

When more customers arrive, the autoscaler starts more instances. A new instance needs about 10 seconds to start. Until it is ready, it serves nobody.

So you need enough instances ready before the peak, not after it.

You also choose CPU and RAM for each instance. More CPU: each instance serves more customers (about 120 per second for a full vCPU, a bit less on time). More RAM: a busy instance can keep customers waiting in a queue instead of losing them.

At the peak760customers per second
One instance48customers per second
Ready at the peak16instances
  1. 1. Customers riseThe CPU of your instances goes up.
  2. 2. Instance askedThe CPU passed your scale-up trigger.
  3. 3. StartingAbout 10 seconds. It serves nobody yet.
  4. 4. ReadyIt serves customers, and costs money until it stops.
04Try it

Get the black line above the wave

Turn a knob: drag it, scroll, or use the arrow keys. The wave replays each time.

Always running, even before the wave. You pay for them all the time.
The most instances you allow. It limits what you can serve, and what you spend.
The CPU level that asks for new instances. Lower starts them earlier.
  • No customer lost (no red)
  • Almost nobody waits (amber under 3%)

Loading the wave…

Customers per secondArrivingYou can serveWaitingLostMaxMin
Money lost$0
Lost checkouts0
Cloud cost$0.00

05The score

Lose less, spend less

Each wave has two costs: the money you lose (customers kept waiting or lost) and your cloud cost (every instance, every second, even when it waits for nothing).

Your score goes up when both go down. Each league also has a budget per wave. If one wave of a run goes over it, the run does not count.

Training runPlays every wave of your league and shows your score. Unlimited. It does not count.
Ranked run3 a day. Your best ranked run of the month goes on the leaderboard.
Your last try
Money lost$0waiting and lost customers
Cloud cost$0.00your instances
SCORE---
Cloud budget per wave$7.50Over budget on one wave, and the run does not count.
06Ready

Your turn

That's the whole game. Now practice on the waves of your league: a practice wave replays each time you move a lever, as often as you want.

Open the control room Play: sign in and name your team

You can open this tutorial again from How to play in the control room, or read the full player manual and the FAQ.

Remember
  1. Purple above black means customers wait (amber) or are lost (red).
  2. New instances need about 10 seconds. Have them ready before the peak.
  3. Max instances limits what you can serve and what you spend.
  4. Lose less, spend less, and stay under the budget.
Step 1 of 6