By Antony Melvin – Architecture & Planning Practice Lead
Super-cloud architecture is a logical evolutionary step from the current multi-cloud future that most large organisations foresee. Moving to a single platform that reduces cloud complexity and cost is compelling – but is it likely or feasible?
Multi-cloud or super-cloud?
The multi-cloud concept is well established, utilising aspects of some major public clouds and on-premise servers for the right features. If Google does something better or more innovative than Amazon – then use their service. And vice versa. The downside is that you need to be across several highly complex systems to be able to make that call.
The super-cloud (AKA cross-cloud/meta-cloud) sits above the multi-cloud. A simple explanation of the super-cloud is that it is a service layer that sits above several clouds and offers a single place and terminology (possibly object storage rather than Amazon S3, GCP Cloud Storage or Azure Blob Storage) to manage security, networking, server management, storage and so on.
If you have a programming background you could think of it as an interface describing cloud contracts that are then made concrete by the implementation in Azure, GCP, AWS (Amazon Web Services), on-premises or wherever – and you code to the interfaces. Infrastructure as a high-level language rather than the intermediate language that the cloud providers give you while they abstract away the complexities of the binary involved in spinning up servers and securing them.
For the Site Reliability Engineer or DevOps staffer the implementation complexity of each aspect of the multi-cloud is abstracted away by the super-cloud so that in terms of IT management you only see the higher level.
Building a super-cloud
An interview with David Linthicum, Chief Cloud Strategy Officer for Deloitte inspired this blog and explains why he thinks that the super-cloud will be with us soon. The more thought leaders like him discuss the concept, the quicker it will be that someone takes the hint and releases this cloud super-set.
As a concept it makes complete sense, but the potential roadmap and product management sounds hugely complex. By complex I mean scary. By scary I mean costly. Complex and costly because the number of underlying services that each of the cloud providers are building are exploding. Simply looking at AWS, around 95 services in 2018 has jumped to over 250 services in 2022. Keeping up with AWS, Azure and GCP would be rolling 3 boulders up that hill.
While the worth of a service layer super-cloud seems clear, wrapping every service from every cloud into a higher-level is a task. A more achievable super-cloud would be one that does not try to chase every new cloud service and focuses some core functionality around security, storage, networking and compute. Spin up a server, store some files, log some events. You want it done securely, resiliently, quickly but you are not interested how or where it is done. This sounds like an achievable super-cloud.
I can also see that you would need to use higher level Infrastructure as Code (IaC) tools to write the abstract layer, like Terraform or Pulumi. It would make little sense to use AWS CloudFormation, Google Cloud Deployment Manager or Azure Resource Manager for a super-cloud. Focusing on a subset of services across the clouds is necessary to have the correctly usable IaC components. It has been a while since I did set theory, so I do not remember what to call a superset of a series of subsets. A compromise. Or a product roadmap.
Are multi-cloud devices super?
Some vendors focus on discrete, abstracted multi-cloud use cases, and in the short term that is close to super.
I attended an AWS meet up in February 2020 (days before the full extent of the oncoming storm became apparent). Panzura presented their hardware device that abstracted away storage across the 3 major public clouds and chose the cheapest available. I have never used Panzura CloudFS – but the concept slide made sense. Of the 2 colleagues who went to the meet up with me, one of them now also works at Deloitte. The Cloud Strategy Officer’s thinking about super-cloud prompted this blog. Is that meta?
The super-cloud does seem to be a missing piece of the multi-cloud architecture. If 90% of large organisations see their future as multi-cloud then expecting to be able to recruit line staff to be simultaneous experts in AWS, GCP, Azure, on-premise, whatever – is clearly unicorn nonsense.
I think that there must be a focused super-cloud based on a tight MVP ready to go somewhere. Even if expert multi-cloud architects look like a horned horse to me – hopefully, super-cloud architect are a more achievable future.