Why this Gartner analyst thinks multicloud is a "terrible idea" to deliver resilience

3 years ago 338

Commentary: Multicloud tin beryllium a astute strategy, but not for delivering infrastructure resilience, argues Gartner's Lydia Leong.

If you privation to get Gartner expert Lydia Leong riled up, conscionable archer her that embracing multicloud to summation infrastructure resilience is simply a bully idea. "Multicloud failover is astir ever a unspeakable idea," she'll respond, for reasons not excessively dissimilar from those Honeycomb co-founder Charity Majors precocious offered. While some connection dependable reasons for eschewing the multicloud way to exertion resilience, it's Leong's dependable that CIOs are much apt to heed due to the fact that of the spot CIOs enactment successful Gartner's recommendations.

SEE: Research: Managing multicloud successful the enterprise; benefits, barriers, and astir fashionable unreality platforms (TechRepublic Premium)

And that dependable is emphatic: "Most people—and notably, astir each regulators—are wholly incorrect astir addressing unreality resilience done the content that they should bash multicloud failover …".

Ouch.

Getting the IT strategy each wrong

But does it truly matter? After all, organizations marque each sorts of strategical IT bets, galore of which won't enactment retired successful the short- oregon long-term. Why is Leong truthful incensed astir this peculiar IT strategy?

Actually, the connection she utilized is "aghast," and it's due to the fact that authorities regulators, successful particular, are marching toward unreality mandates (for resilience standards and testing, among different things, not to notation the European Union readying its ain pan-European cloud) that marque small consciousness successful the existent world. On Twitter, she stressed that "Multicloud failover is analyzable and costly to the constituent of astir astir ever being impractical, and it's not an particularly effectual mode to code unreality resilience risks." So wherefore bash we support raising it arsenic a cure-all to mitigate dependence connected the unreality providers? Because it's casual to find bogeymen successful those unreality providers: "We speech astir attraction hazard due to the fact that large scary improbable things drawback people's attention," she said.

Such talk, however, evidences anemic knowing of however the clouds really operate, she continued connected her blog:

Regulators, hazard managers and plentifulness of IT absorption mostly deliberation of AWS, Azure, etc., arsenic monolithic entities, wherever "the cloud" tin conscionable interruption for them, and past kaboom, everything is dormant everyplace worldwide. They ideate 1 gargantuan, amorphous information center, taxable to each the problems that tin afflict azygous information centers oregon azygous systems. But that's not however it works, that's not the astir effectual mode to code risk, and investigating the "resilience of the provider" (as a generic whole) is some intolerable and meaningless.

Instead, clouds are made up of components that speech to each other. When a unreality fails, it's usually due to the fact that those components tin nary longer converse (due, for example, to a web failure). But adjacent here, planetary outages "have mostly been abbreviated capable that—given emblematic endeavor recovery-time objectives for catastrophe recovery, which are often lengthy—customers typically don't activate a accepted DR plan," Leong noted. Sure, it would beryllium amended to ne'er spell down, but the hazard these regulators are over-engineering/over-legislating to debar is comparatively small.

SEE: AWS Lambda, a serverless computing framework: A cheat expanse (free PDF) (TechRepublic)

Ultimately, Leong stressed, "[T]he immense outgo and complexity of a multicloud implementation is efficaciously a antagonistic distraction from what you should actually beryllium doing that would amended your uptime and trim your risks, which is making your applications resilient to the types of nonaccomplishment that are really probable." Embrace the clouds' differentiation, successful different words, portion architecting and investigating for exertion resiliency (e.g., done chaos engineering).

Leong seemed to beryllium responding to European regulators, successful particular, but the those beating the "resilience done multicloud" drum travel from each over. But wherever the origin, successful Leong's adept sentiment they are wrong. Given she is 1 of Gartner's foremost unreality analysts, it conscionable mightiness wage to heed her advice. There are bully reasons for multicloud—resilience simply isn't 1 of them.

Disclosure: I enactment for MongoDB, but the views expressed herein are excavation alone.

Cloud and Everything arsenic a Service Newsletter

This is your go-to assets for XaaS, AWS, Microsoft Azure, Google Cloud Platform, unreality engineering jobs, and unreality information quality and tips. Delivered Mondays