We aren’t putting modern applications on mainframes. Nor should we be shirking one cloud provider because another says they don’t believe in the multi-cloud.
By Craigh Stuart, systems engineering manager: SADC & WECA at Nutanix
If we have taken any lessons from the recent pandemic, it is that we don’t buy all our apps from one provider. Instead, we purchase apps for the right job from the right place, and we land up in a multi-cloud era. Whether we initially planned to or not.
If you are running some of your workloads or consume certain services in AWS and you purchase Office 365 which is hosted by Microsoft, you are in a multi-cloud world. There has been this misconception around multi-cloud only being hinged to the notion that you make a conscious decision as IT to buy real estate in multiple cloud environments – where it is much simpler than that. Multi cloud by its nature enables organisations to choose a platform to deliver a specific service or application across various cloud providers. They could be public, private or edge providers that enable a business outcome. The definition is simple – the practice and management are not.
Unfortunately, the multi-cloud is fantastic on paper, but it just isn’t that easy. The problem here becomes governance, complexity, management, data sovereignty and security. None of this is new – we know this, you know this, we aren’t the first to say it. If you map out a multi-cloud architecture and try and create a single view of not just where your cloud real estate lies but how you can pool it together to create a centralised layer as you had in your on-prem environment, it starts getting complicated.
Business must create a very granular view of what a multi-cloud stack looks like, where each component of your business lies, where there are dependencies, proprietary layers, overlap and workload allocation in order to future proof the organisations IT should they need to pull out or scale a component. The “all-in to the public cloud” rush doesn’t make sense, it never did, but it was tried anyway, and customers have been burnt.
But that doesn’t mean that navigating the multi-cloud is a cheap exercise at first. Following your initial investment, your costs will go up before they come down. Because your usage and workload allocation are incremental your costs will adjust as you adjust your model and rationalise your usage to the point where you know where workloads are run most cost effectively.
What the multi-cloud is not
A multi-cloud environment should never be focussed on how your individual cloud service providers factor in your architecture. It should be centred on the software, tools and resources that sit above these clouds and make this model viable to the business. You need to extract the cloud out of the conversation and start thinking about how you configure the technologies you use into your multi-cloud solution – that is where the magic begins to happen.
In the past, there was a level of comfort in being caged in by the on-prem data centre you built and controlled. Most organisations purchased software stacks but ultimately ended up with shelfware. Public cloud is the complete opposite of how we have been consuming IT in the past. It is about consuming services and applications that you need to deliver an outcome and pay for those services.
Start-up companies today don’t go and build their own data centres. They go and co-locate something if they need to do something special with their apps and workloads, or they select a public cloud to run apps in and choose SaaS providers to host their business apps in (think HR and Payroll). They manage IT as a pure-play consumption model, something that the whole notion of computing needs to start moving towards.
Unlock the lock-in
Large organisations can’t still be sitting in a dark room, flipping coins to decide where they are going to be building these big data centres and what they are going to be putting in them. Instead, we are moving to a model where the tooling or clusters that enable management of workloads and applications become more important than the cloud itself.
The cloud really shouldn’t matter, just like the brand on the hardware you use shouldn’t matter. CIOs should instead be considering building a management cluster or control plane that allows them to orchestrate workloads where they want them and when they want them. With this, they can then go an subscribe to what the business needs, this ensures that what you subscribe to becomes Opex based, allows IT to spin services or workloads up or down, makes it utilisation based and opens the way for computing anywhere.
The world has changed. We aren’t shoeing horses before a long trip; we are refuelling cars, the same with IT. Yes, it was a world where everything was centralised in the office, but with the pandemic, everything is distributed with people in pockets all over, some even at home.
But the future question is, where will we be two years from now? If you have forklifted everything into the cloud without considering that we might need to go back to a more flexible version of the old centralised model, how do you bring your data and your apps back closer to the user?
This is precisely where the complex piece of the multi-cloud puzzle lies. Which is why creating a management cluster is so critical today. IT can’t wait to put this portion in, the time to act is now because without a view of where all your disparate clouds and systems are, making any changes to a future operating model isn’t going to be hard – it will be downright impossible.