Recently, I’ve done a quick summary of public cloud usage among customers with whom I had worked before. This was a result of queries that we get from other clients who ask for consultancy in this area. My research confirmed what I somehow instinctively knew all along: the majority (i.e., 65%) of our clients are fine with a hybrid cloud strategy relying on a single public cloud. They don’t need more than one “cloud” to be effective or innovative!
Start With the “Why”
Let’s analyse for a moment why companies are drawn to multi-cloud. Originally, multi-cloud was an approach with which you could host the same workload on different cloud platforms and change the processing location on demand. Since then, the term has been overused, and now it basically means using many clouds.
Having multiple cloud platforms is useful if you’re afraid of vendor lock-in and prices going up in the future. In reality, though, prices are mostly going down because of the scale of cloud operations and innovation.
Another reason for choosing multi-cloud might be the desire to use the “best of breed” cloud services. For example, Google is renowned for its natural language processing services, so if you’re building an application with that for a specific market, you might want to choose the Google Cloud Platform.
The third motive is related to building and selling a software product. If you need to be flexible, e.g., ready to run your solution on a cloud chosen by your client, you have no other option than to build a cloud-agnostic solution. In such cases, you need at least partial agnosticism to minimise your long-term maintenance costs.
The fourth multi-cloud scenario occurs when you have to host your solution locally on different clouds in order to be compliant with certain regulations. For example, at the moment, none of the top cloud providers has a data centre in Peru. So, if you had to build a solution with high focus on data sovereignty for the UK & Peru, you would have to choose “multi-cloud” or “hybrid cloud” paths.
Finally, some systems might require extremely high availability. I’m thinking here about solutions for sectors such as banking, energy or defence. Even in their cases, though, taking this approach is debatable. A more complex solution, which I’m sure a multi-cloud one would be, is not necessarily one with higher availability.
That’s the pretty and alluring theory for applying multi-cloud. What happens in reality is that some companies go multi-cloud by accident. If you have many employees on-board, inadequate IT governance and/or pressing delivery deadlines, I’m sure that at some point somebody will introduce another cloud platform. That’s just how life works.
Nobody Told Me About the Disadvantages!
You might be thinking it’s not that bad. If you were a carpenter, having another tool in your toolbox wouldn’t hurt, would it? Well, it’s different with multi-cloud.
So, what’s more difficult? First of all, with two clouds you have to either hire twice as many specialists, or your employees have to learn twice as many cloud services. In the long term, this is ineffective, and it will limit your innovation potential, as it’s going to be more difficult to catch up with two evolving cloud offerings.
Furthermore, infrastructure management is more difficult, too. Either you do it separately for every cloud (sharing just basic account details like username and password), or you buy a tool to help you with multi-cloud management. This tool, which initially looks like a way out, means an additional cost for your company, even more learning for your employees and a potential lock-in (which you might have tried to avoid in the first place when selecting the multi-cloud path). So, consider all the pros and cons carefully!
Then comes the vendor management. With two cloud vendors you have twice as many meetings, contracts, renewals, and guess what? Your negotiating position for discounts has just been undercut, because now you spend only half of the money on a single cloud.
Finally, tracking the cloud spend and cloud reporting will also require twice as much effort, as each vendor uses different concepts, labels and tools to expose that information.
I hope that now you see that choosing your cloud approach needs to be a conscious and informed decision. I reckon Gregor Hohpe summarised it nicely: “Hybrid (cloud) is a reality, multi (cloud) is an option”. And just as with other choices in the IT architecture, this option doesn’t come for free.
Head of Architecture
He manages a team of over 30 Software Architects. He has 17+ years’ experience in commercial software development and is responsible for supervising clients’ software architecture. Occasionally, he’s also involved in certain key projects and provides technical consultancy services.
He’s a co-organiser of the WROC# conference as well as conducts company training sessions on estimation- and bid-related topics.
This article was originally published on Objectivity’s blog: Why You (Probably) Don’t Need Multi-Cloud