DEV Community

Cover image for Multicloud hysteria - and because (a lot) probably is not for you
Raphael Moura for AWS Community Builders

Posted on • Updated on

Multicloud hysteria - and because (a lot) probably is not for you

Can't happen any problems in some great cloud provider, that my LinkedIn timeline fills with people talking about multicloud, with great taste as if it were the 8th wonder of the world, something easy like getting sweet from the child's hand. Unfortunately, 99% of those who speak of Multicloud have no idea how to start and have no idea of the work to manage.

Multicloud's myth has passed the hand of several "experts" on the market, but none of them tell you the big problems that exist with this. It's pretty good on paper, but there are several difficulties to put into practice and to work 100%. Let me tell you some (many) disadvantages of Multicloud.

Dodge vendor lock-in. But does it make sense?

First, let's understand the concept of lock-in.

Lock-in happens when the cost of changing provider is greater than the cost of permanence or when you are bound not to change due to the incentive that you currently receive from your cloud provider

Often, you enter lock-in because both realities are real when it comes to the public cloud. And exactly where is the problem in this?

Historically there were problems related to high costs? AWS For example, it is a provider that always lowers the costs of your services, it never increases. And this also depends on the cost factor of applications or architecture. If you have a fully serverless infrastructure, it becomes so cheap that it does not justify getting out (companies should take everything to serverless as they achieve, because cost and efficiency justify). A lot is talked about lock-in but there are rare cases where this is a problem within the public cloud because you can get the best advantage of a provider, you get the best cost and the best performance. And you're not wrong to do that.

But let's say you're afraid of it. Then you will have to use Open-Source tools that will give you this power. However, you will still have to manage absolutely everything in 2 different clouds. You will have to take applications that could be serverless for your Kubernetes cluster, for example, and only in it, you double your management cost. Is it still worth it?

Cost savings, multicloud justifies?

Image description

Probably not. For the simple reason of management. You learn deeply everything about a cloud provider is already a difficult task because you have visibility of everything that is happening, manage access, compliance, security, expenses control, deployment of applications, etcetera, are challenges that the IT team will need to do.

But you can only do this because the team deepens your knowledge within a cloud provider. Now imagine for example a DevOps team managing deployment in 2 cloud providers for different applications? There's where the problem lies, and most likely this cost you would already not be worth it. But there is more.

When it comes to providers such as AWS, there are costs for inbound and outbound traffic. For example, if you have applications that are going to connect with a provider X database while the application is in a provider Y, depending on the amount of data this becomes a cost problem, and I do not need to talk about the management of connections, both in Visibility and security.

What ends up going on is that you become good at 2 providers, but not excellent. And it does not take the best of any. All because you want to have multicloud.

But the clouds have outages, I want to prevent it!

Image description

How about starting to build your applications in the right way? When there were problems in US-East-1 with AWS, the fault is from exactly?

Well, AWS should not have these problems, but you should predict this. And if you can prevent this by simply using a different region. Now if your application should always stay up, you have to maintain a contingency in another region to keep your application alive.

But it is easier to talk about the provider than solve the problem. And a little secret, you do not need another provider to lower Blast Radius. And a problem happens globally, for example? Virtually impossible, but if you still want to go to Multicloud, good luck.

And I'll use the best of the cloud providers if I have multicloud?

You can, but the cost will be too high. Many speak of multicloud but none of them use their cloud in the best way. They do not use the best AWS, Azure, or GCP and want to talk about using more than one cloud! When I say it's hysteria is not a joke.

Even before starting with multicloud, you must first know your main cloud provider very well. Know everything about it and get the best of it. Probably when you do this maybe do not even think about using another cloud, but if you use it will do it in the best possible way, and of course, having aware of obstacles of managing and costs.

And when should I use Multicloud?

Image description

When the cost and advantages justify, and when you have the maturity and sufficient knowledge in a cloud provider, use the best of it. It has scenarios yes that you can use, such as a database in a separate provider, or the famous case of Azure AD which for me is well justifiable (but for me multicloud is much more than using only a service from a provider) but should you have understood from what I said here. The biggest advantage of using multicloud is that yes, you leave vendor lock-in, you can use the best of more than a cloud provider, and you can also focus on cost because you will use what has a better value of each. But all this comes with the cost. If it's worth it, and you meet everything that was said, you can be a good deal.

But I believe that for 99% of today's companies, it is not worth it. Who knows when you have multicloud management platforms that can bring us this power, the scenario changes?

Top comments (0)