Recently AWS had an outage at their US-EAST-1, which is troubled site for AWS historically. It was explained as a “database issue that was magnified due to a networking issue”. I don’t believe it— it sounds like a standard BS explanation we have all grown to hear. They screwed up something and are ashamed to admit it or think being open about it will reveal too much competitive information.
These things happen with any infrastructure. This should have been mostly a non event. But most startups running on AWS had given no thought ( probably time-starved) to architecting their app infrastructure. They had put everything in US-EAST-1 without thinking about DR scenarios. Its also due to the tyranny of defaults, since when you sign up, that’s where AWS puts you in. Signs are that Netflix was able to avoid an outage because of the work they have done to architect their environment.
Some people are using this as a reason to create FUD about Cloud. I have seen 3 types of FUD:
1. You must have multi-cloud strategy
Because AWS had an outage, you should have had capability to shift workloads to Google or Azure on demand. Really, people? So, you telling me that these startups that don’t know how to properly architect to take advantage of different regions and zones in AWS will automagically figure out how to dynamically migrate workloads across disparate clouds. Whatever you are smoking, its good stuff, keep at it, because you won’t like it when your trippin’ ends.
2. This is perfect reason for doing Hybrid
Yeah, like Cloud bursting, automagically failover to the server at Uncle Bob’s basement. I won’t discuss this much except to say that computer science can be a bitch for people who don’t take time to learn.
3. Build Private Clouds, so you can control your own destiny
Yes, this actually makes sense and somewhat of a reasonable point to make, except for the fact that 95%+ of private cloud projects fail and you probably can’t afford the army of people and ELAs you need to pull off something like this. In addition, private clouds won’t scale to meet your needs and your growth will suffer and you will end up having a “Zynga” moment.
So, my point kids is to treat AWS outage as a learning experience to ask right architectural questions and tune out vendors that want to sell you moar boxes.