Are all your digital eggs in one AWS basket?

The recent AWS outage in Sydney once more highlighted the business case for a multi-cloud strategy. We talk to Bailey Caldwell of RightScale and Greg Steel from the Offis Professional Services team about the unremitting rise of multi-cloud adoption.

Recently the East Coast of Australia was hit by a huge storm which caused disruptions to Amazon’s Web Services. Many of their customer’s business websites—including Domain, Domino’s Pizza, Foxtel and The Iconic — went offline, remaining unavailable for hours. Yet other businesses were able to continue largely unaffected because they had previously structured their hosting strategies to use more than one data centre or cloud.

Outages happen. They always will. While some commentators suggest this latest event should be a warning against migrating to the cloud, others see it as a lesson to businesses to work with multiple clouds and/or zones. How a business survives such an event depends more on how it structures the technology to extract the benefits of the cloud while ensuring maximum reliability and business continuity.

“Amazon [will] still be the first cloud people use,” says Bailey Caldwell, Director of Professional Services for RightScale—the leading multi-cloud management platform. “[However], a lot of businesses don't want to use Amazon for [reasons] like data residency and geography. And there are a lot of businesses [that] won't put all their eggs in one basket. We're seeing those trends increase.”

Many CIOs view a multi-cloud strategy as too complex or expensive, outweighing the benefits. Yet, most may already be operating some form of multi-cloud ecosystem, but without the strategy and management tools to be truly effective and cost efficient.

Multi-Cloud by Accident or Design

 “I think the world went multi-cloud [long] before people realised that they were multi-cloud,” says Greg Steel, senior consultant and solution architect at Offis. “As soon as you signed up for your first Gmail account, you were multi-cloud.” However, while most individuals have little need to be too concerned about which server stores their emails or which cloud powers a favourite app, organisations have to ensure far greater governance and control (sorry, Hillary).

“Multi-cloud is one of those things that would tend to happen probably either by accident or by there being some compelling business benefit. You may have built a whole bunch of stuff on AWS, but if you choose to use a software-as-a-service platform that [runs] on a different cloud provider, say SharePoint on Azure, then you've become multi-cloud. The difficulty then is managing that multi-cloud state.”

Steel points out that organisations making a deliberate decision to adopt multi-cloud is a much more recent trend. The difference between “accidental” multi-cloud and “deliberate” multi-cloud is a clear strategy; choosing exactly which clouds to use in which configuration to achieve maximum benefit—all while maintaining the flexibility to change that configuration when circumstances change.

How many clouds are too many?

“Globally, cloud management platforms are on the verge of becoming mandatory for every enterprise,” says Caldwell. “Cloud is a reality for everybody and the need to deal with hybrid environments is the next step in the adoption cycle.”

Caldwell has watched with interest how multi-cloud strategies have evolved and adapted over the years. “When I first started at RightScale five years ago, [it supported] Amazon and Rackspace. My first tour of duty in the company was building out our partner ecosystem. Over the next couple of years, we ended up working with a hundred-plus partners. We worked with clouds that don't exist anymore. FlexiScale, have you heard of them? That’s so long ago very few people even know they exist. IDC Frontier, who got acquired out of Japan, used to have a cloud.”

However, after a flurry of new cloud services entering the market (and some dropping out of the market a few years later), clear trends began to emerge. “We realised that there are different classes of cloud services,” says Caldwell. “The only class that's really proven itself to be viable is the mega clouds. So, you have companies like Amazon, and Azure, and Google that have enough critical mass to build a 100,000 square-foot data centre, not to serve as cloud but to serve their core business. Those are the only three companies with the capital to exert on this public cloud space problem. I don't think there's going to be a fourth.”

“Then there are the specialised clouds as we define them, which would be a Rackspace with its higher service levels, maybe an IBM SoftLayer with some enterprise capabilities, bare metal. Maybe there'll be a HIPAA cloud, for example. Maybe there’ll be industry domain-specific clouds. So, while we still support eight clouds, I really think of it as the three mega-clouds we support.”

Multi-cloud strategy

While a multi-cloud strategy has become largely unavoidable for most businesses—particularly when trying to maintain a competitive advantage—there are still many questions to answer. How to manage multiple clouds while ensuing each is available, accessible and integrated correctly into the organisation? How to ensure this complex mix performs well enough to achieve the agreed goals? How to maintain or even improve upon the service levels people expect? These and other questions can make it harder for organisations to make a confident, informed business decision about which clouds to use.

“The biggest challenge is that anything you're going to do with cloud [will] involve having systems work with each other; having one server and one cloud over here work with a server that's in a private data centre or a cloud over there,” says Steel. “That sounds simple but to make that happen there are so many different people and different sub-organisations involved, all of which have their own rules and standards and processes … [that] need to get on the same page, pulling in the same direction. Most of the changes … can be challenging or can change their role to an extent that they're not comfortable.”

Ultimately, managing this new ecosystem requires good governance. It shouldn’t be compromised further by the restrictions or incompatible definitions of the management platform. “Years ago, [RightScale was] very focused on this thing called a server template and making a virtual machine a SQL Server or a MySQL server,” explains Caldwell. “We forked an early version of Chef. We did a ton of work with bash scripts and PowerShell and configuration management approaches. That's now an industry unto itself [with] Salt, Ansible, Chef, Puppet. Back then, it was Puppet and that was it. Chef was a brand new player on the block.”

With so many alternative approaches to configuration management now available, RightScale realised it needed to be compatible with whatever made more sense for their enterprise clients. “We've extracted our platform further away from having an opinion about how something becomes a SQL Server [to] caring much more about the overall continuum and the governance factors. You can serve and use RightScale, [or] you can use Ansible, Salt, Chef, or whatever.”

Fewer challenges, more benefits

“The way I like to think about RightScale is this concept of universal cloud management, from bare metal to containers and everything in between: provisioning Platform-as-as-Service; provisioning Rancher or Docker; clusters around containers; as well as virtual machines,” says Caldwell. “It's really the ecosystem you're operating within. Add to that self-service, plug-ins, and all the other HTP services out there that you may want to interact with.”

Steel agrees that universal cloud management has made it easier to solve the various technical, procedural and governance challenges. “A lot of organisations end up seeing either that it's not so scary— it is quite manageable—or that the value to the business is just so huge that you have to embrace cloud and multi-cloud.”