IBM bets on a multi-cloud future

As Amazon, Microsoft, and Google battle for cloud dominance, IBM is taking a different tack. Yes, it is building is own public cloud, and is on the homestretch of harmonizing its disparate Bluemix and SoftLayer cloud flavors. But it is also looking at the sweet spot of its customer base, and what it sees is organizations that tend to use many clouds. It recently commissioned a survey with BCG and McKinsey that quantified what's probably common knowledge: over 90% of enterprises are using multiple clouds, with two thirds of them using two (and usually more) public clouds.

IBM's cloud message is not that its public cloud is bigger and badder, but that it is targeting the management of multiple clouds. And those clouds are going to be running in a mix of environments: inside the firewall, outside in "bare metal" single tenanted private clouds, and public cloud. IBM terms it "hybrid cloud and multi-cloud."

The draw of public cloud is the opportunity to simplify and transform IT, but the risk is cloud vendor lock-in. If you use any PaaS (Platform as a Service) or SaaS (Software as a Service) offering, you are making a vendor platform decision all over again. Even if you simply go to using IaaS (Infrastructure as a Service), you'll have to pay the ransom for extracting any data that you have stored in case you want to take your business (and data) elsewhere.

IBM senses the opportunity to provide the management, security, and governance buffer that reduces the cloud provider lock-in risk. IBM may have missed out on the public cloud building boom that Amazon and Azure exploited, but it has lots of enterprise clients who need to clean up the resulting mess.

IBM has always had lots of enterprise clients with lots of IT complexity, and now that complexity encompasses cloud. It's the mess that happens anytime technology walks in the back door via departmental budgets that bypass IT, or when it becomes easy to trigger new services but hard to keep track of the ensuing mess. It's an old saga that began with PCs, later extended to LANs, web sites, VMs, new line of business applications, and most recently, cloud.

And while IBM is not well known for public cloud, it claims to have the largest base of "bare metal" single tenanted cloud implementations in the industry. That makes sense given that bare metal is the logical outgrowth of IBM's hosting and services business of yore.

Dealing with IT complexity is IBM's core DNA. That was formalized during the Lou Gerstner era, where instead of breaking the company up after its "near death" experience of the early 1990s, he chose to keep the company together and focus on integration. That drove growth of IBM's global services business until it hit the wall roughly a decade ago.

IBM's multi-cloud management strategy makes sense given its history of managing IT complexity. Just as most enterprises wound up collecting one of everything when it came to compute, application, and storage platforms, today they often have one of every cloud. In most cases, this is not because of any formal enterprise strategy to avoid tying destinies to any single cloud provider. Instead, most enterprises wind up with multiple clouds because of inertia.

Nonetheless, IBM can't throw the same old formula in addressing complexity on this go round. There is backlash today from enterprises that are no longer willing to pay the same high multiples for services. Traditionally, if you spent a dollar on infrastructure or software, you often spent several times that for integration and implementation. And anyway, isn't the cloud supposed to reduce IT costs?

IBM faces several challenges here. First, the onus is on IBM for rolling out tooling that won't automatically lead to heavy services engagements. But secondly, it must ensure that the cure for complexity is not more complex than the disease. It must address questions as to what level of management and oversight solves rather than compounds the problem.

For instance, will it be necessary for IBM to duplicate the function of Amazon CloudWatch when the specific client's point of pain centers on controlling costs? And by the way, yes, IBM has a brokerage solution just for that. Then there are some functions, such as optimizing the choice of the right compute, storage, and network instances for the job that will be best left to the cloud provider, which will have the most current information on their portfolio.

On the product side, a couple prime examples of IBM's multi-cloud include its cluster management and private cloud platforms. IBM Multicloud Manager is a tool for monitoring and managing Kubernetes environments, pinpointing component and service health, and managing applications across clusters and clouds -- plural. It includes capabilities for setting security policies and workload management. IBM's value proposition is that providing a cross-cluster, cross-cloud view becomes more manageable than handling it individually, within each cloud or cloud service. It's an ambitious remit, much akin to having the pane of glass in the sky for application and systems management in the enterprise.

IBM Cloud Private (ICP) is a self-contained cloud software platform that bundles and integrates open source frameworks, like Kubernetes and Cloud Foundry, with common services for deployment, monitoring, and security. IBM is coming out with three flavors of ICP: one for data and analytics, another for integration, and another for applications/business processes. While ICP has been billed as private cloud, IBM is now supporting the option for customers to run ICP on public cloud through IaaS. In other words, use Amazon EC2 as the place you physically deploy, but use IBM as your cloud environment.

While ICP is about cloud provider independence, the fact is that it is an IBM platform. In any IT procurement decision, at some point must commit to choosing a primary vendor. The advantage of ICP is insulating from the dependencies of any particular cloud platform provider. But traversing that route has its limits: you won't be able to take advantage of cloud provider native serverless offerings like AWS Lamba, Azure Functions, or Google Cloud Functions. Instead, you'll have to rely on IBM's packaging of the open source Apache OpenWhisk.

Of course, the penguin in the room is IBM's pending acquisition of Red Hat. IBM is billing Red Hat's OpenShift container management platform as the nexus for turning each cloud into a neutral plane with portability. The confusion was where ICP picks up and OpenShift leaves off, as they are both Kubernetes container orchestration platforms. The answer? OpenShift replaces the container orchestration platform that IBM originally built into ICP; otherwise, the rest of the ICP stack (encompassing IBM middleware along with common management and security services) remains unchanged. If you have 18 minutes, IBM will be glad to explain it in this video.

Given that IBM was not first to building public cloud, going to its core DNA of managing complexity is the logical path. But IBM must make the case that stitching multiple clouds into a coherent management plane won't translate to introducing more complexity or writing blank checks to consultants. Maybe the tag line for IBM's multi-cloud strategy should be, clouds were supposed to simplify IT, but when you have a lot of them, nobody said it would be easy.