If giants such as Netflix and Spotify don’t see much value in adopting a multi-cloud approach, why are so many smaller firms so keen on the model?
Without the mavericks who are prepared challenge received wisdom, the business world would be a lot less dynamic – or interesting, for that matter. Take Paul Graham, co-founder of startup accelerator Y Combinator, for instance. Renowned in Silicon Valley for his ability to pick a winner, he recently admitted that he’d never read one business plan or balance sheet supplied by firms seeking an investment from him.
Graham’s justification? “The reason I don’t care about business plans is that I can learn more from five minutes of interrogating the founders than I can from reading the 10 pages of fluff they’ve written.”
And consider fintech unicorn Bolt, which broke its industry’s norm of long working hours by adopting a four-day week in September 2021. Acknowledging that it was an experiment, the firm’s founder and executive chairman, Ryan Breslow, explained at the time: “People are done working like cows for five days. They are ready to work like lions for four.”
There are contrarians in the cloud computing world too. When the consensus overwhelmingly favours hosting applications using a multi-cloud solution (the use of two or more computing services from any number of vendors), a small minority resolutely back the single-cloud model.
The benefits of a multi-cloud approach
But why? The multi-cloud model offers clear advantages for users. For instance, it makes it possible for them to haggle on prices and select the best-value vendor for each service required, because some providers are better than others for certain tasks. In essence, it gives users the choice and modular flexibility that a single-cloud approach clearly cannot provide.
Of the 1,700 IT decision-makers polled in a survey for US cloud computing firm Nutanix in September last year, 83% agreed that a hybrid multi-cloud approach was ideal.
Do the refuseniks have a convincing argument? Is the single-cloud model worth sticking with if no one provider can offer the best value for all services?
Scott Riley, founder of independent consultancy Cloud Nexus, is proud to be a contrarian. He blasts what he sees as three widely held myths associated with the multi-cloud approach.
“The cost differences between the main hyper-scale cloud providers are not that significant unless you’re prepared to make a multi-year commitment,” he says. “You have the price or you might buy through a distributor for a 10% discount.”
The concept of using multi-cloud to geolocate data is also wrong, argues Riley, who says: “All hyper-scale providers have multiple geographical data centres with individual fault-tolerant platforms.”
It’s a fair point. Where in the world is beyond the reach of the major cloud hosts these days? The Pitcairn Islands, perhaps?
Then there’s resilience. Is it really correct to say that multi-cloud is inherently better in this respect?
“Load distribution has to happen somewhere,” Riley says. “Take an inbound request and pass it to the nearest, or least busy, server in any given cloud. Where does this sit: cloud A, cloud B or a third cloud? Wherever it sits, whenever that platform has an issue, no one is getting to the application, regardless of how many cloud platforms you’ve deployed it to.”
Security is one of the biggest aspects of cloud strategy. There’s a widespread belief that using several providers is a more resilient approach than relying on one.
But that’s another popular fallacy, according to Tim Erlin, vice-president of strategy at cybersecurity company Tripwire.
“There are good reasons to have a multi-cloud strategy, but security isn’t usually one of them,” he says. “The security advantages are dubious at worst and require significant investment at best. Most often, the security benefits are ascribed to the protection that multi-cloud gives against distributed denial-of-service [DDoS] attacks, which is really a rehash of the resilience argument.”
Achieving resilience requires a substantial outlay – and there are better ways of doing so than adopting a multi-cloud approach, Erlin adds.
“Building a multi-cloud infrastructure that allows for seamless failover [the facility to switch automatically to a back-up system] across providers requires specific investment. It’s not simply an emergent quality of having multiple providers,” he says. “This might be a worthwhile investment for businesses that require that level of availability and/or are at a high risk of DDoS attacks, but it should be considered alongside alternatives that might mitigate the risk at a lower cost.”
Maybe the biggest argument in favour of a single-cloud strategy is the simplicity it offers.
David Liddle, senior cloud security consultant at Adarma Security explains: “In addition to insider threats, cloud misconfigurations are one of the biggest issues facing users. These misconfigurations, which range from overly permissive policies attached to identities to poorly configured security groups, can be introduced in several ways. Spotting them in a single hyper-scale environment can be extremely difficult. With a multi-cloud approach, the problem of eliminating misconfigurations becomes even more difficult.”
Then there are the risks of migration. Moving from a single host to multi-cloud is a serious undertaking. It requires virtual machines (VMs) to be stripped out, because cloud hosts, while similar in functionality, have their own quirks. A VM that runs on Amazon Web Services (AWS) may not run on Microsoft Azure, for instance.
Companies need to be sure they’ve got the right monitoring tools in place to run a multi-cloud spread. Every extra element is a point in favour of the simplicity of the single-cloud approach.
Why the tech giants avoid multi-cloud
Perhaps the most convincing evidence is from the tech giants. Netflix, for example, runs computing and storage exclusively on AWS. It trialled a multi-cloud operation briefly in 2018 before committing to a single provider. Netflix is no ordinary customer – at one point it accounted for 15% of all global internet traffic. If it had felt that using only one provider were too much of a risk, it surely would have diversified by now.
Spotify prefers one provider too, having migrated from AWS to Google Cloud in 2016. The company’s vice-president of technology, Tyson Singer, said last year: “There’s simplicity in having a single cloud. It saves us a lot of hassle and complexity.”
The approach that both Netflix and Spotify have adopted is unfashionable nonetheless. Multi-cloud is the default option. But it is useful to know all the arguments for and against.
“Is multi-cloud advisable? The answer is far from straightforward,” Liddle says. “For some, multi-cloud is essential. For others, it’s an aspirational money pit.”
Smaller companies in particular may feel emboldened to keep things simple. And, if they do, they can take comfort in pursuing a strategy that suits a pair of digital behemoths.
Trends don’t last forever in business. Contrarians act and followers emerge. Running a single-cloud strategy may feel awkward, but you may simply be ahead of the curve.