Sometimes, the point of a revolution isn’t immediately clear. Though he later claimed to have been misquoted, a Chinese leader once warned us that the significance of the 1789 French Revolution was still “too early to say.” As James Burke will tell you, it often takes quite a while for the impact of change to fully ripple out (“There are times when an invention is technologically possible and it may appear necessary, but in the absence of the technical and social infrastructure to support it, the invention will not survive”).
I wonder if we might look back on Cloud in the same way—as a breakthrough which took a while for the business benefits to become clear.
Anyone who thinks the Cloud is now fully stabilised is wrong. Gartner recently published an analysis of the top Cloud Infrastructure and Platform Services. It emphasised that the sector is still evolving and that new players continue to come through.
Gartner voiced something that, up until now, you might think was taboo; no cloud is perfect for everything. The report insists a multi-cloud strategy – (the use of cloud services from multiple public cloud providers for the same set of IT solutions or workloads) is now the smart play. And asserts that “a well-governed multi-cloud strategy can improve access to a breadth of technology choices and innovative best-of-breed capabilities.”
Just three cloud providers hold over 70% of the market—but it’s naïve to think there’s such a thing as ‘just’ an AWS, ‘just’ an Azure, or even a GCP shop. Certainly, the CFO would prefer it that way, as she figures that’s just one annual subscription to pay.
‘One cloud to rule them all’ may have been the plan five years ago, but it’s now clear that you should tactically use cloud for certain applications, in specific regions of the world. This will protect your data, keep your business nimble and ensure you are never subject to the whim of any one big CSP (Cloud Service Provider).
I mention data deliberately. Here, the issue is data sovereignty as any international business needs to keep on top of different data sovereignty requirements in different regions. No single vendor can meet all of these for you all the time, so you will have to work with different cloud providers to be 100% compliant with local/regional privacy regulations. For true multi-geography data sovereignty, you need different support levels from different providers in different sovereignty regions.
Another driver for multi-cloud is the avoidance of concentration risk. This is a problem that regulators like the Bank of England have specifically called out (see its new draft supervisory statements). If you’re not a Tier One Bank you may think this isn’t relevant to you, but no-one can put all their main business systems in one platform and expect it to never fail (at this point, SME business users will now be shuddering at the name ‘Rackspace’).
Why they invented PaaS… and why you should be sceptical
Avoidance of concentration risk, data privacy, and system crashes aren’t the only reasons to commit to a multi-cloud strategy. Vendor lock-in is also a problem. The CSPs have known from the start that the basic infrastructure as a service proposition, IaaS, wouldn’t be enough to keep the chairman in yachts. Platform as a service, PaaS (where they keep selling you add-ons to keep you from leaving the showroom) is where the real profits lie.
So, the real reason we wanted cloud is finally becoming clear: to keep your independence or manoeuvrability open. This makes multi-cloud the best strategy.
Unfortunately, there is one last ‘Big Boss’ at the end of this cloud level video game to beat. And it’s a big one! The database.
Applications technologies in the cloud are scalable, portable, and open. This is where Kubernetes is king: modern cloud micro services. Open technologies like this make it compelling to work in the cloud, making it straightforward to develop, test, deploy, extend, and scale any business app you want. But your database is the final frontier, as you simply can’t work with it portably.
That’s because for serious business or transactional applications, pre-cloud business systems like Oracle have just not evolved into a cloud database that works as well as cloud native technologies like Kubernetes, Kafka, or Slack, tools you can just pick up and drop in.
Do you want to throw away the whole point of even trying multi-cloud?
People try to address this challenge by trying to chunk up their big monolithic database, or instead try NoSQL or other non-standard business data engines, which just don’t work. (Don’t sue me, it’s just the reality.) Feeling you have no choice; you then revert to the version of a database your eager new PaaS CSP friend wants to sell you, and immediately throw away the whole point of even trying to go multi-cloud.
Maybe we have to accept that if neither NoSQL nor older pre-cloud databases deliver the true cloud independence multi-cloud offers, it must be a chimera. The genuinely good news here is that there is (finally) an option for delivering true multi-cloud multi-database workloads.
That comes in the shape of open source, distributed SQL, where the decades of experience and features SQL offers are all wrapped up in an easily accessible form, like PostgreSQL. Even better are properly productised versions that really align with the scale and ACID issues of transactional work--a prime example of this is our own YugabyteDB.
I’m not saying the cloud revolution is going to follow the path of the French Revolution (for a start we don’t need Madame Guillotine). What I am saying is that the smart move with cloud going into 2023 is informed by Gartner’s report; cloud is a dynamic market whose innovations are far from done and choices continue to come to market. To maximise your business options, multi-cloud should also mean considering adopting distributed SQL to give you the best possible options for business success in the Cloud.