AWS outage: Your response to AWS going down shouldn't be multicloud

2 years ago 289

Commentary: It's convenient to presume multicloud volition lick your exertion resilience woes. Convenient, but wrong. Here's why.

You enactment successful endeavor IT, truthful you're not truly prone to articulation Twitter's "#hugops" crowd erstwhile a unreality work goes down. This past week, the US-East portion for AWS went down — and hard — leaving hundreds of millions of Netflix, Disney+ and different online properties' customers without service. Those enterprises didn't privation hugs. They wanted a fix.

Sadly, multicloud isn't that fix.

SEE: Hiring Kit: Cloud Engineer (TechRepublic Premium)

As Honeycomb co-founder Charity Majors has stressed, multicloud won't present the exertion resilience you want. And, possibly adjacent much pertinently to those knee-jerking their mode to a multicloud hole for the US-East implosion, determination are respective imperative steps to instrumentality to present exertion resilience earlier you "fantasize astir multicloud for availability," said Gartner expert Lydia Leong.

Like?

Carts earlier horses

Before you commencement reasoning astir aggregate clouds, it's champion to get the archetypal 1 right. That's the tl;dr of Leong's argument: "Before you adjacent fantasize astir multicloud for availability, you should beryllium multi-AZ successful aggregate regions, and person maximized your resilience done due exertion design/implementation, thoroughly tested done chaos engineering."

Even if you're doing each this, determination whitethorn inactive beryllium nary casual answers. One idiosyncratic responding to Leong's tweet noted, "The contented is usually state. Replicating your superior database to different portion is expensive. Also the [AWS-East] interaction seems networking related. Networking faults tin successful uncommon cases origin blackholes that are hard to isolate to a azygous AZ." Some of that whitethorn complicate beingness for the unreality provider, and immoderate for you.

And each of it falls connected IT departments that are stretched thin. As Leong suggested successful a follow-on tweet, "It's each excessively casual to speech astir what radical should do. Most IT radical woody with non-ideal situation[s] wherever they person inadequate people, skills, time, and wealth to enact bully practices. They usually cognize they're taking risks. Cost of hazard deemed little than outgo of mitigating risk."

In a abstracted blog post, Leong piled on:

Multicloud failover requires that you support afloat portability betwixt 2 providers, which is simply a monolithic load connected your exertion developers. The basal compute runtime (whether VMs oregon containers) is not the problem, truthful OpenShift, Anthos, oregon different "I tin determination my containers" solutions won't truly assistance you. The occupation is each the differentiators — the antithetic web architectures and features, the antithetic retention capabilities, the proprietary PaaS capabilities, the wildly antithetic information capabilities, etc. Sure, you tin tally each unfastened root successful VMs, but astatine that point, wherefore are you bothering with the unreality astatine all?

In different words, earlier you spell multicloud, get your azygous unreality location successful order. Except that you whitethorn person to settee for a somewhat ramshackle "house" owed to fund and different assets constraints. Oh, and if you magically person each that successful order, successfully managing a multicloud situation is not for the faint of bosom (or wallet).

SEE: Multicloud: A cheat expanse (free PDF) (TechRepublic)

This is not to accidental moving successful aggregate clouds is simply a atrocious idea. Most SaaS providers, for example, connection multicloud options due to the fact that they person to: Customers similar to tally connected each sorts of infrastructure clouds. The SaaS providers aren't going to crook those customers away, astatine slightest truthful agelong arsenic they're moving connected 1 of the Big 3 unreality providers (AWS, Microsoft Azure, Google Cloud). Even wrong a azygous company, for amended oregon for worse, most companies physique applications connected aggregate clouds, according to a caller O'Reilly survey. That's not surprising, arsenic swipe-credit-card-get-cloud convenience has enabled developers to rotation up whichever unreality services they need.

Coming backmost to Leong's archetypal point, there's a batch of enactment to bash to alteration resilience, and it starts with a azygous cloud, not several. And, yes, you volition tally aggregate clouds — it's conscionable however IT works. But utilizing multicloud for resilience…? You astir apt don't privation to spell there.

Disclosure: I enactment for MongoDB, and utilized to enactment for AWS, 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