Forbes | Five Multicloud Considerations For A Data-Centric World
Last year, I wrote about the imminent multicloud wave. Multicloud has rapidly become one of the most-talked-about IT concepts. IT leaders are now either asking how they should use multicloud for their enterprises or facing the reality of multicloud implementations, and it is challenging. Over the past year, as the CTO of Faction, I’ve worked with enterprises on a number of multicloud initiatives and have seen some patterns emerge.
One surprising truth is that multicloud is less of a revolutionary idea and more simply a nearly inevitable result of enterprises seeking to gain the most agility for their initiatives. Similar to cloud itself, multicloud is not automatically the best choice for any and every workload; it is more specialized than cloud. It shares another similarity to cloud in that it can be tricky to implement. Therefore, many enterprises will defer it if they can even if there is a clear business benefit to adopting.
Here are five considerations about multicloud that are coming up over and over again as enterprises look at adopting multicloud patterns:
1. Data (and its locality) matters.
You can’t ignore data gravity, and it will become a stronger consideration over time. You can solve for many other multicloud challenges with software, policy and planning, but you can’t change the speed of light. Across industries, from biotechnology to finance to media and entertainment, with a wide variety of edge and cloud initiatives, the locality of data and how it is seamlessly accessed in an efficient way is a tricky challenge. This also applies to where you place your cloud workloads because not every region is equal when it comes to major cloud providers.
2. Remember your developers.
Of course, Kubernetes is on the tip of the tongue. With VMware’s announcement of a multicloud-centric Kubernetes control surface in Tanzu and similar efforts from the likes of Spinnaker and Rancher, there is a clear movement to allow teams to easily deploy any containerized app to any infrastructure in an increasingly standard way. Developers will drive this effort because this is largely an effort to help them build and deploy faster and better. Your multicloud strategy must also be data-centric.
3. Unstructured data is the big data challenge.
Data is everywhere, and multicloud is inherently concerned with it. But unstructured data is larger and growing faster than structured data. This, combined with the challenges of moving data, poses an architectural challenge for applying multicloud usage to big data problems such as machine learning, and we already see efforts to solve for this in some of the most data-centric parts of the industry, such as connected cars.
4. There are many reasons to want more than one cloud.
While goals like resiliency, financial leverage and avoiding lock-in get a lot of talk, we have been finding that other things are driving true multicloud projects. We’re finding features of HPC hardware such as newer, faster GPU cards, as well as availability and scale of spot instances and, most notably, the “unlocking” of data from a particular cloud, are key drivers.
5. Edge applications are not fully defined.
The era of edge computing is arriving with fanfare and is poised to disrupt the technology landscape. We can see twin forces of edge and cloud executing a data tug-of-war that must sift data between “urgently available” data and “must be centralized” data. With use cases like connected cars, moving data in from and out to the edge are both critical.
The exciting thing about using a multicloud architecture is the appeal of not choosing. If you’re adopting cloud for agility and innovation, do you want to rule out a big portion of that agility and innovation just because it isn’t coming from your first choice of cloud provider? Of course not, but without a valid multicloud strategy, governance and architecture, you may not have a choice.
Remember that when it comes to hype, after the peak of inflated expectations comes the sobering disillusionment of making things work as an early adopter. That doesn’t mean you can rest on your laurels because enterprises rely on IT to help them win, especially when conditions turn rough. Thinking early about the data an enterprise generates and how it can help them win helps bring clarity to a multicloud strategy.