3 ms·
AWS has different risks. Many, such as Linode, will bend over backwards to fix your systems if they've suffered a hardware failure or data corruption issue. In
by astrodust 13y ago
AWS has different risks. Many, such as Linode, will bend over backwards to fix your systems if they've suffered a hardware failure or data corruption issue. In using them for six years, with 50-60 instances at any given time, I've only permanently lost one instance.
AWS is a purely do-it-yourself model, where hardware will blow up from time to time and data will get lost. Any and all infrastructure on AWS is to be considered temporary. Safety only comes with redundancy or a service layer like EngineYard that handles durability and persistence for you using their tooling.
In three years of using raw AWS and hosting about a dozen instances, I permanently lost six instances, had several others corrupted and rendered unusable. You need to be prepared for parts their infrastructure to completely evaporate at any given time. Everything has to be persisted to S3 and off-site relentlessly.
As much as it's true that you need backups for any provider, it's especially true for raw AWS. You're a tiny boat in a big ocean of trouble if you only have a few instances, no redundancy, and infrequent backups.