Which cloud should we use?

CloudBank supports four clouds: AWS, Azure, Google Cloud Platform and IBM cloud. How do I figure out which one works best for my research team?

They All Work Great

Before digging in to this, we would like to make what we see as a reassuring observation: For many research programs, any of these cloud platforms will work great. The cloud providers actively maintain feature parity; and while the specific details vary the basic services are very equivalent. After all Linux is Linux 99% of the time, regardless of which Virtual Machine it runs on.

But Let’s Break This Question Down A Bit

Cloud choice for research is a slightly involved topic, as one might imagine: All research projects do have their own unique character. So going in, a good ‘meta-question’ to keep in mind might be “Does our research program benefit from a careful cloud intercomparison (which can become a bit of a project in itself)?”

The CloudBank portal features a catalog of features broken out by vendor; a place to start (but caveat: These catalogs are subject to becoming dated as providers release new features.)

Pending further “Cloud Differentiator” content: For the moment, let’s describe a couple of approaches one can take to understanding cloud choice. And remember please: If you are accessing the cloud through CloudBank you have access to all four clouds with no obligation; so you can try them out on an exploratory basis and you can mix-and-match or switch clouds as you see fit. However to minimize your cloud “process” it is ideal to settle on a particular cloud, set it up for use, and get on with your research.

Usability A good way of getting a feel for the cloud interface is to look into free online introductory training. If you can spend even thirty minutes learning about a cloud interface (the browser-based “console” or “portal”) you will start to develop a sense of what that particular cloud will be like to use.

Cost Cloud providers are in competition, primarily for enterprise subscription. For this reason the cost of their base services is often comparable. There are some ‘third party’ cost comparison blogs on the internet (here is an example) that can help illuminate how vendor costs compare. However as a good first approximation we see no clear “cost winner” for basic cloud services (compute power, storage, networking) so it might be more productive to focus on applicability of the cloud stacks to your research. And note the ‘they all work great’ remark above.

Case Studies It can be very helpful to look at case studies: What team X accomplished on cloud Y for topic Z. Web searches can work well for this; and we also invite researchers to write to CloudBank asking for pointers or references for a particular type of research. At this time there is often no substitute for direct communication with (start with an email to help@cloudbank.org) to get some helpful guidance.

There is much more to this important topic; hopefully these remarks will serve as a good starting point.