2 ms·

This article is interesting and the author has some good points about the lack of small non-burstable instances in AWS. There are plenty of things to gripe abou

This article is interesting and the author has some good points about the lack of small non-burstable instances in AWS. There are plenty of things to gripe about. But the headline assertion is ultimately unsupported except in terms of simple comparison graphs of paper numbers about the raw CPU and memory numbers. Are the CPU units comparable? Is the networking what it's cracked up to be? How easy is it to autoscale? Is the capacity you need available when you need it? How long do instances take to start? What are the dynamic storage options? How does disk IO performance compare? I'd be really interested to read an article that attempted to break these down and do a real comparison. But this article doesn't even attempt a real world comparison.

user5994461 10y ago
> Are the CPU units comparable? Is the networking what it's cracked up to be? How easy is it to autoscale? Is the capacity you need available when you need it? How long do instances take to start? What are the dynamic storage options? How does disk IO performance compare? In order. Yes. Google networking is superior (cheaper & faster). Variable, depends on your application/workload, not just the cloud. Yes. < 30 seconds on Google, 1-3 minutes on AWS. lcoal SSD, remote SSD, or remote HDD on Google VS a mess of many complex & expensive disks on AWS (it would take more than a blog post to explain their disk offerings). Google disks are 3-10 times IOPS and/or bandwidth, lower latency. This article is just on basic pricing. I plan more articles for the future. Starting with one on disk benchmark and one on network benchmark.