Why one cloud outage takes down half the Internet
Availability zones: how a cloud region survives losing a whole data centre
A storm can knock out a whole data centre. So how does a cloud region keep its apps running? Availability zones, and the strange morning when they did not help.
-
1
Your app lives in us-east-1, and a storm there could stop it. But look closer: the region is not one site. It is six separate groups of data centres, many kilometres apart, each with its own power and its own network. Each group is called an availability zone.
-
2
So you never run just one copy of your app. You run three, in three of those zones, and each ticket scanned at the dock goes to one of them.
-
3
If a zone loses its power, its copy of your app stops with it. The two others go on answering, and at the dock, Rono never notices. Spare copies like these are called redundancy.
-
4
That is what makes the morning of 20 October so strange. No fire, no flood: the three zones stood fine. Yet the three copies of your app failed in the same minute. What did all three share?
-
Quick check
A shop runs two copies of its app, in two zones of one region. A fire stops one of the zones.
Show the answer
Yes. The copy in the other zone keeps answering: each zone has its own power and its own network.
-
Go further
How far apart the zones are, and the names you see for them.