I'm really curious about creating a fault-tolerant application on AWS. It feels challenging considering how many global services rely on the control plane located in us-east-1. Can anyone explain what's stopping AWS from resolving the Single Point of Failure (SPOF) issue? It seems odd to me that they'd say it's too complex when they're making over $100 billion in revenue from AWS alone. Many of us are heavily invested in AWS, and moving to another provider is not really an option. Are there any plans to gradually tackle this issue? I just heard that the cost of a recent outage could be in the hundreds of billions.
6 Answers
It's funny seeing all the outrage when many users experienced no outages because they didn’t rely solely on us-east-1. If you’re using global services, these outages shouldn’t be surprising anymore; you've probably seen this happen several times in recent years. Instead of just complaining, focus on building resilience in your systems!
Don’t worry, they’ll just roll out their AI fixers to handle it. But seriously, achieving 100% uptime is impossible, and that's likely why they aren’t focused on resolving this. It's also worth mentioning that it might not be a true SPOF, just a critical dependency in their system.
It's important to note that a robust workload can still operate even if the control plane goes down. The real problem arises when the data plane fails. In fact, we used AWS during this outage and weren’t affected at all, despite having a low seven-figure annual bill.
I refer to AWS as "Absolutely Woeful Service." That pretty much sums up how I feel about them lately. Word of mouth still holds a lot of weight, even in tech, and I don't hesitate to share that with people when AWS comes up.
The craziest part is that AWS had promised to address this issue a couple of years back after the last big outage. It seems like they missed addressing significant vulnerabilities. I hope they release a post-mortem explaining what went wrong and how they plan to resolve it, but the vendor lock issue really complicates things. There are migration services out there, but they come at a huge cost.
Honestly, you're right about the vendor lock situation. If customers can't leave, there's less motivation for AWS to prioritize fixing these issues. They're not under pressure to change if their clients are stuck with them.
That's interesting! So you weren't relying heavily on the services from us-east-1? It sounds like that strategy really paid off for you.