According to Pingdom, Amazon S3 uptime for 8/9/2009 was 32.84%! Downtime today is already at 12h 44m. That’s pretty disturbing. I ran a google news search and found nothing related to this outage: http://news.google.com/news?q=amazon%20s3%20downtime&oe=utf-8&rls=org.mozilla:en-US:official&client=firefox-a&um=1&ie=UTF-8&sa=N&hl=en&tab=wn
Here are some links that were forwarded from my colleague, Len.
Regarding a PayPal outage: http://www.businessinsider.com/outage-costs-paypal-users-at-least-7-million-2009-8
Regarding an Amazon bug with international shipping: http://news.cnet.com/8301-17939_109-10304824-2.html
Regarding previous outages: http://www.datacenterknowledge.com/archives/2009/07/19/outage-for-amazon-web-services/
You can see the diagnostics for yourself here: Pingdom – Web site monitoring for 100% uptime. Measure your downtime..
Why do I care?
As a web developer and a web host, uptime challenges are an ongoing concern. People bandy about terms like 5-nines uptime like it’s an industry standard. The reality, as demonstrated by those with the deepest pockets, is that despite our best efforts Internet fabric is more brittle that we would like to tolerate. For example, when I read about Amazon’s S3 and related AWS services I was pretty sure that geographical load balancing was one of the perks. However, if that were true, how could lightning strike cause any meaningful downtime.
What Is A Cloud, Anyway?
A physical server is tied to hardware. CPU fails, and the OS is toast: all the services running on the machine fail. As the services move to virtual machines, what is changing? Certainly there’s a decoupling of the OS from the physical hardware. Virtualization does this and we’ve been happily running VMWare for years. But at what point does the VM become a cloud? I had imagined that a cloud was always distributed from any single VM. I suppose that I’m wrong about that. Does anyone else have ideas on what makes a cloud a cloud and not just a virtual environment?