Interesting analysis for sure, but I think the only solid takeaway here is that (unsurprisingly) Google's private backbone between regions offers a huge performance advantage over the public Internet between EC3 regions.
As the analysis stands, the individual IO stats aren't meaningful. EC2 instances offer a range of I/O throughput in different instances: a t1.micro has 'low' I/O, an m1.small has 'moderate' and an hi1.4xlarge offers 1.1 GB/s write and 2.0 GB/s write. Cost has to figure into a benchmarked analysis here.
And CPU. One of Google's big claims at launch was that CPU time was notably less expensive than on EC2. I'd love to see some data here and clicked the link expecting to find that analysis.
As the analysis stands, the individual IO stats aren't meaningful. EC2 instances offer a range of I/O throughput in different instances: a t1.micro has 'low' I/O, an m1.small has 'moderate' and an hi1.4xlarge offers 1.1 GB/s write and 2.0 GB/s write. Cost has to figure into a benchmarked analysis here.
And CPU. One of Google's big claims at launch was that CPU time was notably less expensive than on EC2. I'd love to see some data here and clicked the link expecting to find that analysis.