A few slides I used to set the stage for a panel discussion on Application-Oriented Infrastructure at a kubernetes meetup in Reykjavík on the 9th of February.
The questions initially posed to the panel members were; why are we so excited about application-oriented infrastructure, why is kubernetes an improvement over existing cloud technologies, or other container orchestration systems, and finally: are we as engineers and our companies benefitting from these abstractions, and furthermore is kubernetes the right platform to choose if you want those benefits?
a.k.a. how we accidentally abstracted hardware away
from apps, and everybody loves it!
Where does this term come
• From a 2016 Google paper named: Borg,
Omega, and Kubernetes:
Over time it became clear that the beneﬁts of
containerization go beyond merely enabling
higher levels of utilization. Containerization
transforms the data center from being
machine oriented to being application
Kubernetes is Google's 3rd generation
container orchestration system
• Schedules containers to a dynamically sized
pool of worker nodes • Schedules containers to a dynamically sized pool
of worker nodes
• Service discovery
• Zero-downtime deployments
• Instant/Automatic rollbacks
• Scalability of pods
• Scalability of cluster by adding new nodes
• Secret and conﬁguration management
• Storage orchestration
• Batch execution
Trend not limited to Google
(or even containers)
• AWS (traditional cloud)
• PaaS (Heroku, etc)
• AWS ECS
• Microsoft Service Fabric
• Serverless (AWS lambda + AWS API gateway)
Why are we interested?
• Developers: fast, reliable deployments + easily
port applications between environments, without
worrying about hardware
• SysAdmins: Simpler provisioning, easier
administration, homogenous monitoring
requirements, less vendor dependence
• Management: faster product iterations, bugﬁxes,
stronger bargaining positions with infrastructure
vendors => better bottom line
Are we there yet?
• And is kubernetes the “right” tool of choice?
• Is it going to be the “linux kernel” of distributed
• Feels like we’re all building tools to solve the
same problems — but do we like kubernetes
precisely because it offers powerful primitives,
but few opinions?