I’m not a fan of clouds. I bristle when I hear the term used. It has become a fetish in business circles where non-technical managers speak of clouds to mask their lack of knowledge of the black arts of IT, and it’s a fixture in vendor marketing brochures, where it satisfies the dual criteria of marketecture: It sounds technical but means whatever a vendor says it means.
I’m still trying to make sense of the three genres of cloud: private, hybrid and public. The former, a private cloud, really seems to refer to managed infrastructure—you know, like we never had in distributed computing. The latter, a public cloud, remains a vacuous concept referring to anything delivered via a Wide Area Network (WAN) and on some sort of subscription contract: application access, managed hosting, storage, backup services, archive, etc. Hybrid cloud is basically a kluge of private and public: internally managed infrastructure, but with onramps to public services intended to augment internal capabilities.
To be honest, every time a vendor takes me through this cloud taxonomy, I’m reminded of the “Goldilocks” fairy tale—the one in which a not-too-terribly-bright home invader tries out multiple chairs, bowls of porridge and beds, looking for the ones that feel “just right.”
Private clouds may be better managed, but they still carry the labor costs of internal IT. Hybrid clouds attack Capital Expenditure (CAPEX) costs by enabling firms to grow their servers, storage and other infrastructure on a subscription basis—without buying, installing and powering gear on premise. Public clouds let you unplug your servers and storage and relegate all IT to an outsourcing company in the sky.
All that remains is for corporate “Goldilocks” to decide which arrangement feels just right.
The only value I’ve been able to identify in cloud speak, at least when the subject is private clouds, is how it enables us to talk about infrastructure management inefficiency—but without using the word management, which has long been a budget request non-starter. Private cloud enabled a much-overdue discussion of management. That said, things got out of control when vendors suggested that private cloud would make IT so drool-proof that end users could provision their own resources. That’s a scary thought.
Public clouds are where the story goes back to being a fairy tale. I take it on faith and 30 years of experience that any service delivered over a WAN can’t possibly deliver a trustworthy or believable Service Level Agreement (SLA). Try as they might, cloud infrastructure providers or application providers can’t guarantee availability, response time, throughput, transaction completion rates or any other SLA-related metric in good faith because they’re at the mercy of the WAN that connects their services to consumers. There’s nothing they can do when heavy traffic slows networks, network routing algorithms slow networks, hacker intervention slows networks (or compromise data integrity), or network vendor infighting and co-ownership of WAN facilities make network performance unpredictable. And, of course, not every locale is well-served by broadband network facilities.
Hybrid clouds bring into managed private clouds all the problems of public clouds and WANs, but with none of the purported outsourcing benefits. Moreover, there are precious few standards for cloud on-ramps. That locks in consumers to a particular service rather than enabling mix and match provisioning.
Bottom line: Like the “Goldilocks” story, we all need to be cautious of bears. Have you actually checked to see whether the cloud provider you’re considering has a Tier 1 data center, or is the vendor working out of a beat-up rack of eBay servers in the corner cage of some hole-in-the-wall, managed hosting facility?
A little skepticism might be in order to prevent the bears from getting you.