The cost of cloud at the edge of reason

• Telcos seem to be torn two ways on their relationships with the hyperscalers
• On the one hand there are huge advantages to getting properly into bed with them 
• On the other, there are existential worries that they’ll all be trapped in abusive marriages
• But there are also signs that the lock-in problem may be overworked 

Perhaps the most intractable and most resonant debate currently playing out in our rapidly changing industry is the old chestnut summarised roughly as: ‘Public cloud is always marketed as an economy option (compared to private cloud or private IT) but is actually much more expensive.” Is it? 

Normally a deadlocked argument in telecoms can be resolved with honour on both sides by some sort of “It all depends” construction. This particular issue - which in one form or another has been going on for years - is complicated by a few factors and seems harder to sort out (see: Danielle Royston talks telco cloud, Totogi, tennis (and funding).

There may be several reasons for this. First there’s just the obvious complexity of the public cloud and the way its services might be sold. That’s to say, careful calculations are needed to work out what comes with the deal and how that might be valued? Then, running on a loop in the background, there are issues around the growing relationships between hyperscale clouds and telcos which many in the telco camp see as a threat to telco technological independence. 

That concern came to a head this year with AT&T’s announcement of its partnership with Azure under which Azure will assume management control of AT&T’s advanced infrastructure (see - AT&T takes a giant 5G leap to the public cloud with Azure). Many telco voices took issue with the wisdom of this move and fears of hyperscale cloud lock-in for both corporates and partnering telcos went up a notch.

Software refuge

So the questions around lock-in and how to free up your organisation from one cloud to use another (especially if you discover that you’re paying more for a set of public cloud services than you expected) now seems to have become increasingly important for corporates and telcos alike, with products and services appearing which take as their role efficient extraction from what might be seen as a restrictive cloud or multi-cloud arrangements.

One such is a “first of its kind multi-cloud managed Kubernetes engine” called Kubernetes Kosmos, from multi-cloud service provider for startups and developers, Scaleway. (Scaleway offers private data center colocation and infrastructure, dedicated servers for maximum control and a modern and elastic public cloud ecosystem).

“Until now, most multi-cloud adoption was a result of opportunism,” claims Scaleway CEO Yann Lechelle, “picking up infrastructure from one provider that another didn’t have, or in a geography where the other wasn’t present. With Kosmos, we are facilitating the transition to a true multi-cloud strategy, leveraging the same managed Kubernetes solution that developers have praised for its ease of use and performance, and taking it one step further.” 

By using Kubernetes Kosmos, Scaleaway claims developers can attach nodes on any cloud provider or local resource at a fixed price per cluster of €99 per month, regardless of the number of nodes. 

Google, which after all was the inventor of Kubernetes, also appears to be moving towards encouraging ‘proper’ multi-cloud capabilities - presumably because that’s what its customers are demanding. At its annual cloud conference this month, Google Cloud made a general release of a data warehousing service that lets its users tap into data held on a different cloud, presumably where the cost of storage is lower. So the future may not be so much about users being assisted in making a cloud jailbreak to get to lower prices, but being given the ability to shop around their multi-clouds - without penalties and performance issues - to buy specific disaggregated services, just the way God intended. 

This dynamic is, in many ways, a re-run of the old IBM v. Amdahl fracas back in the 1970s when Amdahl made plug-compatible modules for IBM mainframes - in particular it made storage units to which IBM took extreme umbridge and got out its lawyers. In the end Amdahl won and its customers got, one way or another, much cheaper data storage subsystems. Google isn’t going to make that “Disruption 101” mistake today. Isn’t proper competition wonderful?