By
Tenry Fu, CEO and Co-Founder, Spectro Cloud
Kubernetes
seems to have almost unassailable momentum. It has become the de facto platform
for container orchestration, just as containers themselves are becoming the de
facto model for building and deploying applications anywhere, for developers
and in production at scale.
But
let's be honest: isn't this what we have been talking about at every KubeCon
for years? It's time for a reality check.
As
K8s moves further into the mainstream enterprise and becomes a serious part of
the production stack for more and more workloads, it seems that for every
challenge Kubernetes solves, it creates another. In research we conducted over a year ago, 98% of operations leaders
reported challenges with Kubernetes. 87% said it created a new set of problems
for their production teams to manage.
Are
we (as a K8s community) really ready for prime time? To answer the question,
perhaps we need to focus on more on what "prime time" means, or more precisely,
what managing K8s "at scale" is all about. And to us, that means collectively
defining the answers to four big questions.
1. Are we really
all on the same page about what "K8s" really means?
When
we talk about "Kubernetes" we're probably not just talking about the CNCF
project. We may also be talking about an abstract concept that sits in the
tangled field of open source, cloud native, CI/CD, DevOps and infra as code. Or
we may be referring to the sprawling and vibrant open source ecosystem of
hundreds of integrations and projects springing up.
Depending
on who you ask, your role, or who you are buying from (if you are), perceptions
vary. One thing's for sure: the average cloud native app requires a dozen of
add-on packs (think service mesh, monitoring, logging, CD tools, etc.) on top
of K8s to run. And that's what your devs will expect. And while Ops teams may
be running the K8s infrastructure, we must all remember it's the devs that are
deploying that end-user-facing, potentially revenue-generating code.
So,
here's our first challenge: let's agree a fresh, real-world definition of
Kubernetes.
2. Is K8s ready to
fulfil the revolutionary promise of portability?
If "multicloud"
(along with its myriad meanings) and "multi-environment" have been key themes
for recent years, the next chapter of K8s might be more than just a deployment
model - instead it can shape entire business models.
Let's
not forget, the basis of containerization is the promise of true portability
and assurance for a runtime environment, and Kubernetes as the container
application platform becomes the abstraction layer for multi-environment
to enable portability. With more and more organizations now expanding beyond
public clouds and data centers, near and far edge locations (centralized and
decentralized bare metal K8s servers in essence), are about to revolutionize
industries.
But
are conventional edge architectures K8s-ready? Think cost-efficiency with
single-node configurations, thousands of clusters under management with all the
bells and whistles of a production environment - only in the most challenging
and remote location.
Our
second challenge: Make K8s practicable for the edge and bare metal use cases
that enterprises need.
3. Is it really
about the K8s distribution anymore?
If
the first wave of container standardization was K8s, the second one was CNCF
conformant distributions. While in the earlier years opinionated stacks played
a role, the open source community - fortunately - "normalized" itself.
What's
next? We must get to a phase where the focus shifts away from the distro "up
the stack", towards consistent and integrated management of packs, including
GitOps and IaC support with exposable APIs, native functionality from best-of-breed
open source projects, safe and dev-friendly self-service, and an architecture
that enables real scalability everywhere.
And
this is our third challenge: Alleviating the friction between the freedom that
developers want and the control that IT operations need to maintain.
4. Can humans
manage K8s clusters?
For
most businesses, K8s usually starts with a cloud managed service and a lot of
safe experimentation, DIY stacks and a couple of clusters to play with. But in 2022,
moving to production at scale is looking more like hundreds of diverse
clusters, across diverse locations, with diverse combinations of packs on top,
to support diverse dev projects and diverse teams.
You
can't manage this with hand-crafted YAML files, fast fingers in a CLI and
endlessly scrolling log files. You get the picture: we need machines to manage
machines. And this is where today's idea of declarative template doesn't go far
enough. It's all very well adopting a declarative model and defining the
end-state in day 1, but have we really reached a point where container
orchestration is truly automated, clusters are decoupled from the management
plane, can autonomously self-heal and minimize configuration drift and
ultimately application downtime?
So
our fourth and final challenge: we need capabilities for automation far beyond
human scale, desired-state based management all the way, not just a one-time
deployment template.
Let's talk about
what Kubernetes at scale means for you at KubeCon in Valencia
We
could not be more excited to be in Valencia in just a couple of weeks to help
collectively shape the next era of K8s. So, join the conversation (with some
delicious food) at our panel discussion "K8s
after the Honeymoon: It's Complicated" moderated
by The New
Stack's Alex Williams and Heather Joslyn. We'll talk
about the promise of K8s but also engage in a bit of "group therapy"
about the hard realities of today!
We
are also excited to be diamond sponsors at this
year's Kubernetes
on Edge Day, with an amazing day full day
of keynotes, lightning talks and sessions covering
the today and the tomorrow of one of the most challenging use cases and
deployment models for K8s and open source application innovation. Finally,
don't forget to drop by the Spectro Cloud booth (S57) to say hi and see
Palette, our full-stack K8s management platform in action.
Looking forward to seeing you there!
##
***To learn more
about containerized infrastructure and cloud native technologies, consider
joining us at KubeCon + CloudNativeCon Europe 2022, May 16-20.
ABOUT THE AUTHOR
Tenry Fu, CEO and Co-Founder, Spectro
Cloud
Tenry has more than 20 years of
experience in the tech industry and software development. Prior to co-founding
Spectro Cloud, he led the architecture for Cisco's multi-cloud management and
private cloud portfolio, as a result of his previous venture, CliQr
Technologies, being acquired. Past experience includes VMware and McAfee. He
has more than 18 patents in the fields of scalable distributed systems,
enterprise system management, and security. He is a hardcore audiophile and
likes hiking with his family.