has_many :thoughts: The trouble with services, part 2

The trouble with services, part 2

Colin A. Bartlett — Fri, 15 Feb 2008 09:45:00 -0500

In our last installment, during a Basecamp outage, I concluded that occasional service interruptions didn’t dissuade me from using hosted services. I’d rather have someone else running around working to fix the problem then me.

In today’s installment all of Amazon web services are in-operational.

That means all of Web 2.5, as Justin calls it, is down for the count. I first noticed this on WIRE. After I got over the initial panic of not being able to see a picture of Randy’s Cherios cereal bar, I realized Basecamp, Twitter, and a ton more all depend on Amazon.

This brings us back to the debate over hosted services. And I think there’s one component I missed last time: the number cogs in a wheel. It seems like having services that depend on services that depend on services can bring in additional complexity. So now, not only does using Basecamp require all their servers to be running but it also requires the Amazon servers to be running.

We have debated using Amazon S3 and EC2 here on a few projects recently. Andy asked me straight up not two weeks ago how much I trusted Amazon for reliability. I replied: “100%. They’re Amazon.” I suppose I should have remembered that being a big ass corporation doesn’t make them exempt from downtime. And if one truly needs 100% uptime, one needs to use multiple services with redundant backups. That’s what we all thought we were getting with Amazon web services. I guess not.

"The trouble with services, part 2" by Derek

Fri, 15 Feb 2008 12:15:39 -0500

I don't think that all the services depend on each other. I use Wrike.com for project management. It's a pretty good tool, not perfect, but stable. I don't recall any serious outages. Anyway, I do agree, that it's much better to have someone else running around to fix the problem than to do it yourself.

"The trouble with services, part 2" by Derek

Fri, 15 Feb 2008 12:11:35 -0500

I don't think so. If Basecamp depends on Amazon doesn't mean all other tools do the same thing. We use Wrike.com. The tool is not perfect but pretty stable. I don't recall any serious outages. But afterall, I agree with you that it's much better to have someone else to run around to fix the problem.

"The trouble with services, part 2" by Colin

Fri, 15 Feb 2008 11:59:01 -0500

That's a good point, Jon. AWS has the ability to provide A LOT of redundancy (maybe not complete, but...) for pennies. Literally.

"The trouble with services, part 2" by Justin

Fri, 15 Feb 2008 10:26:10 -0500

Oh, AWS is definitely awesome, no doubt about that!

"The trouble with services, part 2" by Jon Henshaw

Fri, 15 Feb 2008 10:15:28 -0500

Heh, even with the outage, I still think the service is awesome. We couldn't afford to do half of the things we do without AWS.

"The trouble with services, part 2" by Justin

Fri, 15 Feb 2008 10:03:59 -0500

These types of systems are only leading up to the big 3.0... I guess we're really in a more 2.2 or 2.3 if that... definitely not 2.0 only more.

"The trouble with services, part 2" by Justin

Fri, 15 Feb 2008 10:02:33 -0500

Which is why we need to start relying more on distributed systems... *ahem*rdf*ahem*...