Hi.
My company would like to study whether to move Automic to the cloud, or outsource it wholly.
Broadcom has a lot of marketing material regarding "Cloud Editions" of Automation Engine. Frankly, much of this is very fuzzy: It's nowhere clear whether the "Cloud Editions" are stand-alone, separate products or not (apparently they are not, I have discovered). Much of it drops buzzwords such as "AI" and "ML", which have nothing to do, strictly speaking, with "the cloud".
Beyond the fuzz, my perception is as follows:
- the Cloud is merely "someone else's computer". An EC2 instance, for instance, is pretty much a spawnable Linux environment that doesn't differ much in it's functionality from having a Linux box in my own data center.
- nothing technical prevents installing and running a plain old, regular Automic AE in, for instance, Amazon EC2, as of today and already for many years in the past. To my knowledge, CA even had Automic on EC2 years ago, as sales demos.
- the Cloud has potential benefits in scalability
- and it has operating costs and matters of responsibility (uptime, server maintenance) that are differing from running an AE in one's own data center
So, apart from that:
Are there any
tangible, actual benefits of running Automic in the cloud?
Yes, the scalability angle looks nice at first. One could raise a new server and worker processes at the press or a button, or even wholly automatically. HOWEVER we assessed in the past that increasing our worker processes beyond the 20 or so we already run only lead to performance degradation, because the bottle neck with Automic for us seemingly is database dead locks, and more workers only meant more dead locks. So yeah - no matter what the Support says: more worker processes = more bad for us. That puts a major damper on scalability.
Then there's also Kubernetes. I'm defintely no expert on this, but K8s seems like a rather overhead-heavy implementation of early BSD jails or tar'ed up deboostrap environments on Debian, where you run stuff encapsulated in it's own little environment that can be spawned and destroyed at will. Seems great for micro services and, again, for scalability. But does it have any actual, tangible benefits for a heavy-weight software such as Automic, with uptimes ideally measured in months or years, and an update every few weeks or months at best?
I spoke to a Sales guy from a reseller, and his verdict was the same as mine. No, no real benefits. Not in the cloud, unless the cost calculation is beneficial, and also no real benefits in Kubernetes.
Is he wrong? What are we overlooking? Why did you move Automic into the cloud / into K8s?
Also: are there any companies out there that operate Automic as a SaaS (or PaaS, the definitions are blurry), i.e. operate Automic for cash, and allow clients to run their own jobs in an installation managed by third party?
Thanks!