What Should You Learn From the AWS S3 Outage
So, what did we all take away from this week’s AWS outage? Depends who you ask. Plenty of us in IT think it was a good reminder to consider multicloud and hybrid cloud environments—basically, do not put all your cloud workload eggs in one AWS basket. The S3 storage outage affected about 150,000 websites hosted in Amazon’s US-EAST-1 region data center, which is in northern Virginia.
The Amazon outage reminded everyone in IT for a day that cloud workloads are off-premises, and that they can go down at any time. Cloud still needs a recovery, backup and continuity plan. There’s also the time-tested caveat that people still run the cloud machines. The S3 outage this week happened when an engineer made a command mistake during a routine task shutting down a storage subsystem.
But there are other ways to think about public cloud provider outages (which are much less frequent than they used to be). Instead of just saying you’ll start researching multicloud deployments or building more hybrid cloud, one industry expert says that designing for failure is what you should consider for cloud infrastructure. Recommended tips to consider right away include choosing AWS cross-region replication and doing testing for outages and recovery solutions. Longer-term tips are to pick the cloud API that’s best for your business and not sign on for a second cloud provider solely for failover reasons.
On the topic of hybrid cloud, this story explores the serverless computing trend—saying serverless is essentially the same as Platform as a Service (PaaS) computing. With that approach, the middleman (procuring hardware) disappears, allowing IT teams to provision resources directly onto cloud providers’ platforms. Eventually, one consultant says, enterprise IT may choose to directly connect back-end services to clients and client applications, so that there aren’t so many stops between users and the cloud functions they need.
Of course, one cloud provider, at least, was able to feel smug when the AWS outage didn’t affect their own cloud, just announced this week. Nimble doesn’t store its cloud volumes in either AWS or Azure because it can’t guarantee reliability. It’s instead built a proprietary public cloud service for block storage. This may be good both for Nimble’s sales and their customers’ stability. It’ll be interesting to see whether this starts a trend.
Till next week, hug those servers tight.