Hacker News Reader

Comment by Hisoka | Hacker News Reader

Imagine launching your startup on the day Amazon AWS goes down? On another note.. why don't these power outages ever affect Amazon.com from going down? Come'n, eat your own dog food!!!

View on HN

Amazon specifically tells you to set up multiple servers in multiple availability zones. They probably follow their own advice, and, as a result, don't go down.

I've talked to some people about AWS about this, and the reason why they have availability zones is because they don't want to charge you the speed cost of syncing data between zones if your app doesn't need 100% uptime. Generalized replication slows down your app. AWS gives you the option of not having replication or bringing your own.

The last two outages have affected multiple availability zones in the US East region. To really account for it, you'd need instances in different regions.

Multiple AZs went down this time. Only way to have survived it would have been to be multi-region which is...harder.

Amazon.com doesn't run on the same, precisely speaking, infrastructure. My understanding from other Amazon employees was that it (EC2) started as an internal tool.

...and only Amazon can afford to follow their own advice. Multiple AZ hosting ain't cheap. Most CEO/CIO/CTO types spit out their coffee when they see the costs of fully redundant hosting in the cloud at which point they decide "For that price we can afford to be down for 24 hours."

Very few businesses need 100% uptime. As long as you have good recovery strategies, and exercise them routinely, you should be set. When was the last time you ran a failover simulation? do your ops guys know what to do? are there clear lines of communication as to the status of the event?

Outages are hard to avoid, but the pain can be lessened if your customers are aware of the recovery progress and you can deliver on your recovery time goals. Nothing is worse than being down, and leaving customers in the dark to start rumors that your guys are not even aware of the problem.

What? You mean without a backup plan to substitute in for the platform with known reliably issues? Yeah I guess that would suck.

Edit re your edit on dogfooding: supposedly they use the same tech, just not the same servers, so their AZs are probably separate form ec2s. Also, they're in all of their geographical locations, not just us east.

Do you have a source for Amazon.com running on AWS or AWS technology? As far as I know, they've never gone on the record with that.

I'm on my celly so I can't, but if I remember after I land ill try and dig up where I read about it.

According to http://en.wikipedia.org/wiki/Amazon_Cloud the Amazon.com retail website has been running on AWS since November 2010.

Wow - looks like you're right! I dug into the presentation (link below, important slide is #33) and Amazon are now claiming that all their web servers run on AWS. It's only their webservers (not their DBs, load-balancers or "services" which I presume means anything stateful). Of course, the co-location option for 'difficult' services is only available to Amazon, and before the switch-over in November they presumably weren't using AWS nearly so heavily, but this is at least a step in the right direction.

http://www.slideshare.net/AmazonWebServices/2011-aws-tour-au...

they still use different AS for their own retail stuff and cloud services. see radb/bgp

Amazon.de went down when the outage started, but they quickly recovered.

Reply on news.ycombinator.com