WeAreDevelopers Logo WeAreDevelopers
Sign in
For You LIVE
World Congress 2026 Europe • Jul 9, 2026 • Session details

The Hidden Costs of CPU Limits in Kubernetes

Pavel Malyarevsky

Are your Kubernetes CPU limits secretly spiking tail latency? Unmask the hidden micro-throttling cycle and learn how to instantly eliminate waste by tuning runtime concurrency.

Pause
Mute Enter Fullscreen
#1 about 3 min

Driving factors for CPU limits in shared Kubernetes platforms

Shared Kubernetes platforms use limits to avoid noisy neighbors, ensure policy compliance, and manage multi-tenant budgets.

#2 about 2 min

Virtual machine core allocation versus Kubernetes cgroup throttling

Unlike dedicated virtual machine cores, Kubernetes requests and limits use cgroups that halt workloads when CPU quotas deplete.

#3 about 2 min

Impact of Linux quota replenishment on application latency

Depleted CPU quotas pause workloads until the next 100-millisecond replenishment cycle, increasing latency significantly.

#4 about 3 min

Why CPU usage metrics obscure underlying container throttling

Standard CPU usage metrics fail to indicate throttling because halted containers stop consuming cycles during quota waits.

#5 about 4 min

Demonstrating thread contention and container restarts with node exporter

Testing node exporter with concurrent requests reveals how liveness probe timeouts trigger container restarts during CPU throttling.

#6 about 2 min

Comparing performance improvements by increasing CPU limits

Increasing CPU limits reduces throttling and request failures but fails to address underlying excessive thread concurrency issues.

#7 about 3 min

Optimizing execution time and throttling by reducing concurrent threads

Aligning application thread counts to actual CPU availability drastically reduces throttling and lowers overall CPU consumption.

#8 about 2 min

How language runtimes influence runnable thread concurrency

Different application runtimes inherently produce varying numbers of concurrent threads that exacerbate CPU throttling under limits.

#9 about 3 min

Configuring Go runtime concurrency with the GOMAXPROCS variable

Limiting concurrent threads via runtime settings directly decreases the speed of quota consumption and operating system context switching.

#10 about 4 min

Strategies for balancing hard isolation and application performance

Platform engineers should leverage CPU limits for isolation while tuning thread concurrency based on specific workload service level objectives.

#11 about 2 min

Key takeaways on aligning runtime concurrency with CPU quotas

Matching internal application concurrency with CPU limits is essential for both custom workloads and default platform operators.

#12 about 4 min

Cascading failures and node degradation from unmanaged threads

Unrestricted application threading splits CPU time into impractically small increments and triggers self-amplifying request queues that crash entire nodes.

Matching moments

48 sec

Key takeaways on Go container performance optimization

Rick Rackow Rick Rackow · World Congress 2026 Europe

3:57 min

Common Kubernetes misconfigurations in production environments

Noaa Barki · World Congress 2022

3:13 min

How GOMAXPROCS behaves inside limited container computing environments

Rick Rackow Rick Rackow · World Congress 2026 Europe

2:05 min

Identifying Go runtime overhead in high-throughput systems

Ivan Sinitsin Ivan Sinitsin · Europe 2026 Virtual

1:10 min

Optimizing Kubernetes clusters for resource and cost efficiency

Christian Grieger Christian Grieger · Europe 2026 Virtual

3:05 min

Addressing scale complexities in Kubernetes environments

Bassem Dghaidi · LIVE

More from World Congress 2026 Europe

Building the Nervous System of AI - Michael Kagan (NVIDIA) thumbnail
29:17
about 2 months ago

Building the Nervous System of AI - Michael Kagan (NVIDIA)

Michael Kagan, Sead Ahmetović

Five Things in Tech that Matter Now - WeAreDevelopers World Congress 2026 Closing Keynote thumbnail
19:41
about 2 months ago

Five Things in Tech that Matter Now - WeAreDevelopers World Congress 2026 Closing Keynote

Christian Heilmann

The Sound of Your Secrets: Teaching Your Model to Spy, So You Can Learn to Defend thumbnail
32:06
about 2 months ago

The Sound of Your Secrets: Teaching Your Model to Spy, So You Can Learn to Defend

David vonThenen

Super scaling for the Super Bowl: How to survive 30 million users hitting your backend in 30 minutes thumbnail
35:13
about 2 months ago

Super scaling for the Super Bowl: How to survive 30 million users hitting your backend in 30 minutes

Irina Branovic

Running Secure Life Science Research at Scale using Hybrid GPU HPC and Kubernetes 🧬 thumbnail
24:23
about 2 months ago

Running Secure Life Science Research at Scale using Hybrid GPU HPC and Kubernetes 🧬

Jeremy Murray

Angstfreude - AI and Corporate Culture - The Thrill and the Threat thumbnail
30:26
about 2 months ago

Angstfreude - AI and Corporate Culture - The Thrill and the Threat

Axel Ebert, Karin Krobath

Upcoming sessions on this topic

Open session

World Congress 2026 North America

September 25, 2026 · 09:00–09:30

Stage 3

Your Thread Pool Is Lying to You — Sizing Concurrency from Rate Limits and Latency, Not Guesswork

Ratul Ghosh, Sesha Chennupati

Ratul Ghosh
Sesha Chennupati
Open session

World Congress 2026 North America

September 25, 2026 · 16:00–16:10

Outdoor Stage

public void saveMoney(AI): The Developer's Guide to Unit Economics

Hrushikesh Pokala

Senior Software Engineer Lead at Equifax

Hrushikesh Pokala
Open session

World Congress 2026 North America

September 24, 2026 · 15:30–16:00

Stage 9

Run your agents in Kubernetes: Build once, deploy anywhere. But really?

Michal Salanci

Senior Systems Engineer at ESET Cybersecurity

Michal Salanci
Open session

World Congress 2026 North America

September 25, 2026 · 11:00–11:30

Stage 5

Managing GPUs by Just Asking, Infrastructure in the Age of MCP

Jessica Garson Beauchemin

Developer Relations Lead, Community at Runpod

Jessica Garson Beauchemin
Open session

World Congress 2026 North America

September 25, 2026 · 15:30–16:00

Stage 7

Trust, But Verify: Continuous GPU Validation at Scale

Kyle Bell

VP of AI @ TensorWave

Kyle Bell
Open session

World Congress 2026 North America

September 24, 2026 · 11:00–11:30

Stage 1

Application-Defined Compute: Rethinking Infrastructure for AI Applications

Anurag Goel

Founder and CEO of Render

Anurag Goel

Related videos

Go's Concurrency and Parallelism Inside Containers thumbnail
30:48
World Congress 2026 Europe • about 2 months ago

Go's Concurrency and Parallelism Inside Containers

Rick Rackow

Kubernetes Security - Challenge and Opportunity thumbnail
42:45
World Congress 2022 • about 4 years ago

Kubernetes Security - Challenge and Opportunity

Marc Nimmerrichter

5 steps for running a Kubernetes environment at scale thumbnail
57:09
LIVE • almost 6 years ago

5 steps for running a Kubernetes environment at scale

Stijn Polfliet

Instant KAI Sandboxes with vCluster: Multi-Tenant, Multi-Scheduler GPU Sharing thumbnail
27:54
World Congress 2026 Europe • about 2 months ago

Instant KAI Sandboxes with vCluster: Multi-Tenant, Multi-Scheduler GPU Sharing

Piotr Zaniewski

Docker exec without Docker thumbnail
29:16
World Congress 2024 • about 2 years ago

Docker exec without Docker

Oliver Seitz

Running Secure Life Science Research at Scale using Hybrid GPU HPC and Kubernetes 🧬 thumbnail
24:23
World Congress 2026 Europe • about 2 months ago

Running Secure Life Science Research at Scale using Hybrid GPU HPC and Kubernetes 🧬

Jeremy Murray

Related articles

From learning to earning

Jobs that call for the skills explored in this talk.