Enterprises get nearer to the app retailer expertise with Kubernetes and GitOps


Commentary: The massive enterprise drawback is not working lots of of apps throughout a number of clouds; no, the large drawback is working the identical app persistently on only one cloud or information heart.

Picture: Lisa Hornung, iStock

We’re used to getting cell apps by means of Apple’s App Retailer or Google Play, however on the subject of enterprise apps, the expertise is dramatically completely different, and dramatically worse. In response to Weaveworks CEO Alexis Richardson, nonetheless, it does not should be this manner. Although we’re maybe a long time away from enterprise software program being made out there by means of an app store-like expertise, we’re additionally a lot nearer to such advantages than we would suppose. The trick, Richardson mentioned, is not actually to finagle multi-cloud deployment of purposes.

No, for many enterprises, more often than not, they’re simply attempting to make sure consistency in utility configuration on the identical cloud (or information heart). Happily, Kubernetes and GitOps have mixed to make this attainable.

SEE: Hiring Equipment: Cloud Engineer (TechRepublic Premium)

Similar app, completely different locations

This wasn’t the primary time I’ve talked with Richardson in regards to the enterprise app retailer expertise. Neither is it a secret he is shared solely with me. The truth is, at a current convention Richardson addressed the subject in his keynote. What I hadn’t appreciated, nonetheless, was that the enchantment of this “enterprise app retailer” is not as a result of clients are clamoring for multi-cloud. Moderately, it is about consistency throughout the similar cloud (or information heart).  “A key factor that you must allow the enterprise app retailer mannequin is once you do a deployment, whether or not you are establishing a cluster or deploying an app or each, you need the outcomes to be equivalent. So that you need to say it is the identical app somewhere else,” he advised me.

Now that does not essentially imply that each single underlying configuration is equivalent as a result of, for instance, if you happen to’re working on AWS, you would possibly use DynamoDB however if you happen to’re working on Microsoft Azure, you would possibly use CosmosDB. “However basically for the top person it is the identical utility, primarily based on an equivalent set of software program parts, they usually run the identical method on the underlying companies,” Richardson mentioned. 

This feels like multi-cloud, nevertheless it’s not. Probably not.

You realize multi-cloud: the thought that you could take the identical utility and magically transfer it from cloud to cloud, or run the identical utility throughout completely different clouds. Some clients do care about this, however for Richardson, there is a a lot tougher drawback to sort out. Specifically, making certain you may deploy the identical utility, or machine studying platform or no matter, repeatedly and once more, identically on the identical cloud (or on premises), on demand.

“That is simple to say,” Richardson mentioned, “however fairly arduous to do.”

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

Similar. Similar. Similar.

In case you’re not in enterprise IT, working code at scale, this in all probability looks as if an odd assertion. Certainly it must be easy to make sure an utility will work the identical if you happen to’re conserving it on the identical cloud/information heart? Actuality is completely different. Say you and I, for instance, opted to arrange Kubernetes on the identical AWS availability zone (Whats up, US-East!), with the identical machine studying parts. Similar. Similar. Similar. And we hope to make that platform out there to an information science staff inside our firm. Within the Apple App Retailer world, we will be working the identical {hardware}, OS model, and so forth., and any underlying variations Apple takes care of. Within the enterprise, regardless of all that seeming “sameness,” we won’t know that the machine studying platform truly is identical until we are able to evaluate the working programs with a configured mannequin.

Pre-Kubernetes, pre-containers and GitOps, we did not have this capacity. Now we do. GitOps, Richardson mentioned, “supplies a constant method for any developer to handle the entire stack securely, appropriately, and mechanically.” 


Picture: Alexis Richardson

With the Kubernetes/containers and GitOps mixture, enterprises lastly have the flexibility to simply set up and function on a number of clusters and clouds. In flip, firms can run purposes and infrastructure when they need (shutting down companies as wanted), whereas automating the patching/updating of apps, like shoppers have with the app shops. And since it is the enterprise, Kubernetes/GitOps additionally ensures compliance primarily based on enterprise insurance policies and verified operations. That is what enterprises need.

Most enterprises, in brief, aren’t madly attempting to maneuver lots of or hundreds of purposes between clouds. No. As an alternative they’re attempting to get lots of of purposes to behave, to be managed, in a single place. They’re combating foundational points like tips on how to improve these purposes, or create a brand new area, or transfer purposes from edge to someplace else of their infrastructure footprint.

And, whereas they might not have an Apple App Retailer expertise fairly but, we’re getting a lot nearer, due to Kubernetes and GitOps.

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

Additionally see