At GigaOM’s Structure event in San Francisco last week we hosted an invitation-only session for GigaOM Pro clients, to explore the growing popularity of multi-cloud strategies in the enterprise.

A hybrid cloud, in which one public cloud (like Amazon’s, Google’s, or Rackspace’s) is used in combination with one private cloud (an on-premise installation of OpenStack or Eucalyptus, for example) is an increasingly well understood piece of any enterprise IT strategy. Multi-cloud is more complex, but possibly also more capable, and includes more than one public cloud provider and/or more than one private cloud solution. We were keen to understand the extent to which multi-cloud is a conscious and strategic decision, rather than an unwelcome legacy or the accidental by-product of decision making devolved down to small teams or departments within a larger organization.

GigaOM’s Mapping Sessions are highly interactive, tapping the collective wisdom of our analysts and other thought leaders in the market sector of interest. This latest session was no exception, with an hour of informed and healthy debate around a set of tightly interconnected issues.

Terminology

As is often the way in an emergent space, issues of terminology raised their head. Whilst I had considered the distinction between hybrid and multi-cloud sketched out above to be pretty clear, others brought different perspectives. Some of this comes through in blog posts from analysts involved in discussion on the day. Ben Kepes, for example, suggests that

“it seems to me that multi-cloud describes a situation that combines resources from a number of different vendors.”

Discussion during the session tended to use the multi-cloud concept fairly loosely. It appears the stakeholders represented in the room have yet to settle on a clear definition, or a clear set of ways in which multi-cloud can and should be differentiated from closely related areas such as hybrid cloud or — as Kepes mentions — the converged cloud offerings of large systems vendors.

Best of Breed

A clear rationale for consciously pursuing a multi-cloud strategy was quickly identified; the desire to use best of breed solutions from a wide range of different vendors. Rather than depend upon the entire solutions stack offered by Amazon, or Rackspace, or Microsoft, or Salesforce, or whoever, participants clearly and repeatedly stated their desire to select the best tool for any given job. Kepes discusses this in his post, and fellow panelist David Linthicum captures some of the concerns expressed during the session, describing the challenges that traditional IT organizations face in managing integration of disparate components.

The power of the API

Early discussion of best of breed components took place at a high level; Salesforce (for example) for all CRM functions, AWS (for example) for all IaaS compute, etc. However, the conversation quickly moved to a more granular level, and a number of attendees began to clearly express a requirement to combine and exploit fairly small units of functionality in order to construct complex virtual applications built up from pieces of a multitude of different cloud services. Multi-cloud, taken to extremes!

In reality, the only way to discover, request, consume and orchestrate these disparate pieces is through the APIs that individual service providers disclose. There is significant scope for greater visibility of the APIs that already exist, a higher degree of consistency in the way that different APIs allow developers to undertake very similar tasks, and a general increase in the number and capability of APIs available from cloud providers today.

Cloud management tool vendors such as Rightscale and Enstratius already interact with the main APIs of cloud vendors such as Amazon AWS. If this API-powered multi-cloud vision is to become reality for all but the most expert, then companies such as these will need to increase the range of API functions that their products understand and support.

Winners and Losers

Discussion began at a high level, considering multi-cloud as something akin to not putting all of your eggs in one basket, but quickly became far more specific, and far more operational. Conscious decisions around best of breed solutions at a surprisingly granular level, powered by ever-richer sets of API capabilities and a fundamental rethinking (as David Linthicum describes) of the way in which enterprise IT architectures are planned and managed.

The biggest opportunities are for companies developing and supporting the management of cloud resources. Rightscale, Enstratius and others began by offering fairly rudimentary ways to manage the early cloud infrastructure offerings from AWS and others, but have evolved as the market has matured. At least on the basis of the issues raised during this mapping session, there is a clear opportunity for them to continue to evolve, and to directly address the API-powered best of breed requirements that the enterprise is beginning to express.

Slightly further out, there is an equally significant opportunity for a new generation of niche cloud service providers. If the enterprise begins to work through the shift that attendees discussed, and if cloud management tools rise to the challenge, then the pressure on new entrants to deliver an AWS-like range of services significantly lessens. Rather than having to do everything, it becomes far more feasible to do one (or a few) things extremely well. If customers can select your innovative image processing algorithms, or your market-leading fast file storage, or your comprehensive multi-currency billing engine with a simple API call, why would they care that you never bothered to develop an email service, or a hosted Hadoop solution?

Early losers are likely to be the converged infrastructure providers that Kepes described. If best of breed is really of growing significance, then the sellers of all-encompassing hardware and software solutions face an increasingly stiff challenge. Their recent attempts to add value (and increase lock-in) by bundling hardware, software and services will relatively quickly become an impediment to future growth. It will take them time to unpick the pieces and begin selling components once again. Taken to its logical conclusion, the same argument holds true for the larger cloud providers. AWS, for example, is frequently not the fastest, or the best, or the cheapest cloud-based solution, but it is almost certainly the most comprehensive. It also adds new features at a rapid pace, and its breadth, depth, and rate of growth make it a compelling proposition. As customers increasingly select the components that they desire via multi-vendor dashboards powered by API, Amazon and other providers that rely on scale rather than quality will need to think very carefully about the rate at which they cut prices and add truly new features.

Feedback

We welcome your feedback on these disruptive trends. Have we missed anything that you believe will be key to shaping this market over the next two years? Continue the discussion by leaving a comment below.

Look for a Sector RoadMap report in the coming months that will crystallize our take on the Multi-cloud market opportunity.

Panel members David Linthicum and Ben Kepes have written about their perspectives on the discussion, and as moderator I documented some of my thoughts ahead of the session.

Mapping Session panelists

• Paul Miller, Founder, Cloud of Data and Analyst, GigaOM Research
• Ben Kepes, Founder and Principal, Diversity Ltd. and Analyst, GigaOM Research
• David Linthicum, SVP, Cloud Technology Partners and Analyst, GigaOM Research
• Jo Maitland, Research Director, GigaOM Research