So you think all clouds are the same?

Choosing the right cloud provider is not a simple task. There are many service providers offering very distinct architectures. How do you choose? Do you favor one provider because you know their product and are more comfortable with them? Or do you choose based on budget constraints and who can offer you the cheapest services? Both are reasonable criteria, but have you considered that a VM on AWS may perform different from an identical VM on vCloud Air?

vCloud Air vs Azure vs AWS

Earlier this year, a company called Krystallize Technologies stacked vCloud Air against the competition and the results were overwhelmingly positive for VMware. Their CloudQOS platform measures performance across different cloud architectures for capacity planning and thus arming consumers with the facts to select the right solution. When VMware shared the results of Krystallize’s testing, I was confounded. According to CloudQOS, vCloud Air was able to perform 1.6 times more calculations per second than AWS and .95 times more than Google Cloud. You can read about it here.

vCloud Air-competitive-workload-calculation-contest

Independent testing:

The published results of the Krystallize tests were great, but I wanted to see this for myself. I spun up Server 2012 R2 VM’s on vCloud Air, Azure, AWS, and my own test cluster. Each VM had 4 vCPU’s and 16GB of memory (14GB in Azure’s case since 16GB wasn’t an option).

To see the note click here.To hide the note click here.
If you want to try for yourself, I used an m4.xlarge for my AWS VM and a D3 for my Azure VM. My test cluster is made up of two HP BL460 Gen8 blades with two Xeon E5-2690’s and 98GB of RAM.

The first test was a race to see which service could render a 3d image the fastest. I’ve used Cinebench in the past while testing vSphere 6’s SMP-FT and decided to use it again in here because it’s a great way to visualize the processing power of each VM.

Each service did well, taking anywhere from 1:40 to 2:24 to render a single image. vCloud Air came in 2nd place with a Cinebench score of 358, just shy of my on-prem’s 398. A very respectable score nonetheless.

The second test was performed using PassMark’s PerformanceTest to benchmark each VM’s CPU horsepower and memory performance. I was  pleasantly surprised to see vCloud Air outperform my own test cluster. The CPU results below are the overall CPU mark, made up of integer math, floating point math, encryption, and compression performance tests to name a few. The memory test results are also an overall score based on read & write speed, database operations, and latency.


Comparison Table
All monthly costs calculated using current price guides at the time of writing.

What was most surprising was the range in cost among these providers. As expected, running a VM in-house was still the cheapest option, but when placing these workloads in the cloud, the numbers are stacked in favor of vCloud Air. Since vCloud Air outperformed the other providers, you could theoretically get away with a smaller VM to get the same performance and drive costs down further. If you think it’s too good to be true, test for yourself by joining vCloud Air today and claim your $300 in service credits. Be sure to share the results of your own testing in the comments below!


Matt Bradford