Stop adopting multicloud to achieve application resilience, says Honeycomb’s Charity Majors

Commentary: Multicloud could also be good technique in some cases however not for delivering software resilience. Here is why.


Picture: GettyImages/da-kuk

For enterprises nonetheless trying to find the holy grail of multicloud, Honeycomb co-founder and CTO Charity Majors has a suggestion: preserve wanting. Or, somewhat, do not. 

To be truthful, Majors does not imply that multicloud is not helpful for software portability (although Gartner has), or that it could possibly’t be a terrific instrument in contract negotiations (Duckbill analyst Corey Quinn will assure you that it’s not). No, Majors’ level is merely that multicloud will not yield “magical pixie mud” of resilience. Put extra bluntly, she declares, “it’s NOT the way in which to repair your reliability issues.” 


SEE: Analysis: Managing multicloud within the enterprise; advantages, limitations, and hottest cloud platforms (TechRepublic Premium)

When multicloud works

However first, it is value remembering that regardless of all of the hype round multicloud, it is principally simply how enterprise IT works and all the time has. If one nice promise of cloud computing was to unshackle builders from cumbersome {hardware} procurement processes (“Sure, I would like a server for my software. Can I get that by subsequent 12 months? Please??”), that freedom has led builders to purchase into a spread of clouds to entry cloud-specific companies to construct their functions. 

That is what I name “incidental multicloud.” It simply occurs.

There may be additionally “intentional multicloud,” however the place I primarily see that is from distributors (together with the place I work now, MongoDB) who want to fulfill their prospects wherever they occur to be. If a buyer involves Confluent and needs to run the Confluent Cloud on Microsoft Azure, it is a poor gross sales technique to inform them, “No, we solely assist Alibaba” (or no matter cloud). So distributors with cloud companies might assist multicloud to make sure prospects can use information saved in several clouds to run a single software (maybe storing information in Cloud X for price/different causes however analyzing that information in Cloud Y because of superior analytics companies, with out pushing the client to manually transfer the information). 

However that is the seller taking good care of that heavy lifting for the client to make sure a seamless infrastructure/database/software/no matter expertise for the client. That is not what Majors calls loopy. For patrons who assume the burden of multicloud administration within the hopes of better resilience (“I am going to string my app throughout a number of clouds in case one goes down”), Majors desires to speak you out of it.

It seems that the trail to security is not elevated complexity.

Making laborious issues even tougher

“I perceive not wanting a single level of failure. However whenever you add a cloud you do not get extra reliability; you nearly definitely get much less,” Majors noted. Why? As a result of including complexity would not simplify issues. Majors added: “As a substitute of worrying about AWS being down a couple of min a 12 months, now it’s a must to fear about AWS, GCP, and the unholy plumbing between them.” 

This appears apparent, but apparently it isn’t. To not everybody, anyway.

In reality, she went on, it is the plumbing that creates probably the most issues, “since you are not higher than AWS or GCP at constructing and working techniques. Promise.” As talked about above, there are corporations who do have the assets to make sure seamless operation of specific infrastructure/companies between clouds (persevering with the instance above, Confluent guarantees it could possibly “automate[] constructing and monitoring information pipelines and streaming functions, whereas offloading the operational burden of your builders”), however they do not purport to be generalists delivering resilience and different advantages regardless of the workload. They’re specialised in order that IT needn’t be.  

Multicloud might nicely assist with availability points, nevertheless it’s not likely a method for attending to improved resilience–not with out assist anyway. There’s merely an excessive amount of that may’t simply be predicted within the connections between the clouds (Majors: “[I]f you are treating [multicloud] like a sizzling failover, you’ll not take a look at it usually sufficient to forestall nuts & bolts flying off every time”). Once more, in case you have a vendor with the experience to deal with these connections for you, positive. However attempting to run generalized workloads throughout clouds by yourself is not a recipe for resilience–it is a recipe for plenty of resilience whac-a-mole. 

Disclosure: I work for MongoDB, however the views expressed herein are mine alone

Additionally see

Recent Articles


Related Stories

Leave A Reply

Please enter your comment!
Please enter your name here

Stay on op - Ge the daily news in your inbox