As I attended my first-ever Tech Field Day event – Cloud Field Day 13 (CFD13) – in Santa Clara, CA last week, the gritty cult music video Elvis Is Everywhere from the late 1980s kept running through my head. (In this grainy clip, Mojo Nixon and Skid Roper contend that the real secret of life, the universe and everything is that we all have a little Elvis Presley inside us and we’re all moving towards a perfect state of Elvis-ness because of Elvis-lution.)
And if you replace “Elvis” with “Kubernetes,” then you’ve got a glimmer of how I see the current state of DevOps and advanced computing today after attending CFD13: Kubernetes is everywhere, and everybody is using it – even when it may not necessarily make perfect sense to do so. It was amazingly enlightening to sit down with 11 other professionals from across the globe in a hybrid three-day event as we heard from eight different vendors – some huge, some a bit smaller – about the challenges of managing Kubernetes (usually abbreviated to K8, if you’ve been trapped under a virtual rock like me and didn’t already know that) in hybrid cloud, mono-cloud, and on-premises computing environments.
BTC Is Ubiquitous. (No, Not That BTC)
For those of you who know my background already, you probably realize why I was like a fish out of water for the first few hours of CFD13. Every vendor presenting their solutions focused on the Big Three Clouds – Microsoft Azure, Amazon Web Services (AWS), and Google Cloud Platform (GCP) – or BTC, as I term them. And no, we didn’t talk about Bitcoin ($BTC) at all, unless you consider our post-event podcast on the evils of cryptocurrency and its relation to Web 3.0 as relevant.
It wasn’t until the last day that our final vendor presentation from Fortinet briefly acknowledged that yeah, Oracle Cloud Infrastructure actually existed and that their tools could help manage K8 apps in that public cloud environment just as well as any of the BTCs. Yes, Oracle does indeed have a public cloud, it’s pretty damn robust, and in many cases it’s significantly cheaper to operate than the BTCs, especially when there’s considerable data egress volumes, the oft-unspoken-of slayer of DevOps budgets when a developer or QA tester accidentally issues a query that quietly pulls several decades of your company’s sales history across the network. (Stepping off my mandatory ACE Director soapbox for now.)
Like Your New Summer Intern, K8 Demands Care & Feeding
How much care and feeding? Generally, that depends. The vendors that presented their wares to us aligned their approaches to effective K8 management across three magisteria: storage, networking and virtualization, and infrastructure and governance.
Storage. NetApp talked about their ONTAP offering that provides cloud-based storage for K8 applications, Pure discussed their PureFusion product that provides storage-as code, and KastenIO showed off their K10 offering for policy-based data management.
Networking + Virtualization. VMWare talked up their Tanzu application platform as their centralized solution for handling all aspects of K8 security, networking, and connectivity, and Metallic IO demonstrated how their Data Management as a Service (DMAAS) offering as a solution for monitoring the security of their K8 environments at multiple levels.
Again With the Whiteboard
Infrastructure & Governance. I present frequently at Oracle User Group events every few weeks, and I’ve found that relevant use cases resonate the most with my audiences. The folks from RackNGo impressed me the most of any vendor as they highlighted features of the latest release of Digital Rebar. Their demo focused on a not-uncommon conflict: the grizzled insider who’d already built their K8 infrastructure versus the newcomer CTO with an attitude of “I know I’m the new guy, but I’ve got this great vision for our computing infrastructure you are gonna love!” with his prized whiteboard always at the ready. Their interaction reminded me of an aging Captain Kirk’s complaint in The Simpsons’ Star Trek XII: So Very Tired parody: “Again with the whiteboard.” Check out their video on the CFD YouTube channel for the play-by-play of that use case.
Another crucial aspect of K8 management is testing out exactly how K8 environments can handle expected workloads, and the folks at StormForge showed us how their Optimize Pro and Optimize Live toolsets helped predict expected performance vs. live application performance. Finally, Fortinet demoed in real time how their offerings fared against some real-world security incursions against their (I am not making this up) their Damn Vulnerable Web Application.
But Can the Plumbing Take It?
What really surprised me was how little discussion there was about the underlying infrastructure – the databases that power all these K8 clusters, the physical network components and firmware that facilitates flexible virtual networking, and the SSDs, storage arrays, and storage networks that provide retention for massive data sources. During my four decades in the trenches as a DBA / application developer, I’ve been keenly aware of all that infrastructure and how my SQL code, physical and logical data models, and even storage I/O rates affected my applications’ responsiveness.
Our vendors’ presentations seemed to focus completely on enabling K8 DevOps / MLOps activities with scant regard for those “old-school” concerns. Look, I get it: K8 is completely OSS, and our CFD13 presenters are providing some desperately-needed governance tools. But there’s still a a part of me wondering if this K8 craze is so focused on enabling massive scale-out and extremely rapid development cycles when it just might focus a bit more on what seasoned IT professionals know already works: writing efficient code, designing well-formed data models, and giving at least a passing thought to the pounding those physical infrastructure layers are likely to take when we ignore proven IT development methodologies.
