Amazon Web Services Downtime Counter | Hacker News Reader

Amazon Web Services Downtime Counter

awsdowntime.com

38 pointsyuvadam28 comments

How on earth is this number actually being calculated? I'm an AWS customer and all of my ec2 EBS instances are up and running - and that's in us-east-1b, which is IIRC where this outage occurred.

Your us-east-1b is not other customers us-east-1b. Availability zones are permuted randomly for each customer.

Nonetheless, just because you haven't been affected, doesn't mean other have not been, or still aren't.

I really wish they would have picked another way to deal with the fact that people write crappy software that only launches in "1a" and "1b".

They dealt with the resulting capacity imbalances by randomly remapping AZs for new accounts, but AFAIK they're never publicly acknowledged this so most people aren't aware. Instead they should have just started charging more for resources requested explicitly in those zones until people fixed their poorly written software.

They've explained how it works from the beginning.

Right, I understand that. But it seems irresponsible to suggest - as I think this site does - that the entirety of AWS has the uptime you're advertising.

"Availability zones are permuted randomly for each customer"

This statement is false. AZs map directly to physical facilities. If you and I both ask for instances in us-east-1b they will have lower ping times than if one of us has an instance in us-east-1c.

Disclaimer: I used to work for AWS.

According to http://docs.amazonwebservices.com/AWSEC2/latest/UserGuide/us... "Availability Zones are not the same across accounts."

Well then I was wrong. Apologies.

If you're correct, that would be _very_ interesting. The AWS forums are full of statements to the contrary, I thought from AWS employees. The status updates never mention the particular zone (I presumed because the names don't match for different people.) The FAQ says this:

"Q: How can I make sure that I am in the same Availability Zone as another developer? We do not currently support the ability to coordinate launches into the same Availability Zone across AWS developer accounts."

http://aws.amazon.com/ec2/faqs/#How_can_I_make_sure_that_I_a...

If AZs map directly to physical facilities, how does that FAQ make sense?

We're straining to read the tea-leaves here, rather than AWS just telling us what they've built.

Is the percentage right?

AWS went down about 30 hours ago, and in the Seattle area we're 111 days + 7 hours into the year. so by my calculations, it's closer to 98.9% than 99.66...%

Or did I make a mistake?

The calculation is an inclusive, yearly, uptime percentage. This is assuming no other downtime will occur during 2011.

Additionally, this percentage is adequate for the 365-subsequent-days credit calculation. It doesn't matter at which 365-day window you look, my calculation assumes no other downtime has occurred, nor will occur.

Got it. If that's what you're reporting, then you might want slightly different language: "have been up" is past tense, so implied (at least to me) that you were just talking about year-to-date.

Fair enough, will rephrase. Thanks!

For the purposes of claiming service credits from Amazon, the percentage given on your page is not necessarily useful. Amazon's uptime guarantee applies to the trailing 365 days. You can claim those service credits at any time.

Thus, if Amazon had an incident 2 months ago and have another incident now, then your own window is the range [now - 1.year, now] and you figure that this calculates to a 99.8% downtime for you and claim your credit. Your percentage then necessarily goes to zero because your window restarts. But for everyone else who doesn't claim a credit, the window continues to roll forward.

I applaud the notion, but the implementation is atrocious hyberpole:

- simple javascript algebra that does not monitor or report any status data?

- equating the entirety of AWS w/ one region of EBS-backed EC2 instances is hyperbole that does not contribute to the discussion.

How do you guys come up with great web-designing skills even for these simple web pages?

I used blueprint in this case. I have a good history with it, and it always gives me clean, crisp results.

Attention to typography itself can lead to some pretty compelling designs. As Cameron Moll once said, great designers use typography/text as a UI element.

I'm missing something - hours since the start of the year: 2671. Hours of uptime: 2641.

That lead to a downtime calc of 98.88 or so.

*edited for language clarity. Gah! Original sentence above implied that downtime was 2641.

Picking the nit: 2641 hours of uptime from the start of the year until the downtime began, so the uptime calc is 98.88.

it just feels like 2641 hours of downtime since it started :-)

One might point out that "AWS" isn't down, EC2 in VA is entirely broken for a certain notable section of customers.

I'm presuming this only includes downtime since the latest outage began?

Ironically this page is down for me.

Probably DNS hasn't propagated yet. You can use http://goo.gl/rY6cz

That works, many thanks.

urgh, why not make something useful.

I'm AWS! I'm cloudy, I'm bursty, and I'm rocking two nines!

Reply on news.ycombinator.com