Call for submissions: Platform Engineering Meet-ups
Rootconf
Rootconf

Call for submissions: Platform Engineering Meet-ups

Share your journey of building platforms that power engineering teams

All submissions

som

@som23x

How we @Elastic built Multi-Cloud Multi-Region Serverless Platform with K8s

Submitted Feb 19, 2026

Session type: Talk (30 mins) Submission made for: Mumbai meet-up in April

Describe your session in 2 paragraphs

In this session, I walk through the real engineering journey of building Elastic Cloud Serverless on Kubernetes, and the lessons we learned while operating Elasticsearch at massive scale around 600k+ containers with homegrown orchestration with different cloud providers. We start with how Elasticsearch traditionally ran as a large StatefulSet with tightly coupled compute and storage, and why that model limited scalability, resilience, and cost efficiency. From there, I explain the architectural shift toward a stateless, serverless model, decoupling compute from storage, leveraging object stores, introducing dedicated indexing and search tiers, and redesigning how refreshes, shards, and Lucene segments work in a cloud-native environment.

The talk then dives into the practical challenges of running this at hyperscale across cloud providers. We explore key design decisions such as choosing Kubernetes over in-house orchestration, using Crossplane for multi-cloud consistency, adopting Cilium for advanced networking and security policies, solving cloud capacity constraints, and building multidimensional autoscaling beyond HPA and VPA using custom controllers. This session is not theoretical, it’s a behind-the-scenes look at how we made Elasticsearch truly serverless and production-ready across AWS, GCP, and Azure.

Takeaways

  • True serverless requires architectural change, not just autoscaling. Decoupling compute from storage and redesigning refresh, shard, and indexing behavior was essential to making Elasticsearch stateless and cloud-native.
  • Kubernetes alone is not enough at scale. Custom controllers, advanced networking (Cilium), proactive capacity planning, and multi-cloud abstractions were critical to achieving resilience and operational efficiency.

Who is this session beneficial for?

This session is especially valuable for:

  • SREs and Platform Engineers running large-scale Kubernetes workloads
  • DevOps, DevSecOps, and FinOps professionals managing cost and resilience
  • Cloud Architects designing multi-cloud systems
  • Engineers working with Elasticsearch, ECK, ECE, or other stateful distributed systems
  • Anyone curious about how “serverless or stateless” is actually built under the hood

If you’re building or operating distributed systems on Kubernetes, this session provides real-world architectural insights you can apply immediately.

Speaker Bio

I’m Someshwaran Mohan Kumar, a Developer Advocate at Elastic and an open-source enthusiast who grew through community-driven collaboration. I enjoy building projects, exploring codebases, troubleshooting real-world problems, and continuously learning new technologies.

At Elastic, I lead and support developer advocacy in India, working closely with opensource communities to help developers and platform teams adopt and understand low-level designs, cloud-native technologies, K8s, o11y, search, machine-learning and Elasticsearch. I share practical insights through talks, demos, workshops, and technical blogs, with a strong focus on making complex concepts approachable and useful.

All submissions

Comments

{{ gettext('Login to leave a comment') }}

{{ gettext('Post a comment…') }}
{{ gettext('New comment') }}
{{ formTitle }}

{{ errorMsg }}

{{ gettext('No comments posted yet') }}

Register

Hosted by

Rootconf
Rootconf
We care about site reliability, cloud costs, security and data privacy