65k nodes on GKE, with Maciej Rozacki and Wojciech Tyczyński
Guests are Maciej Rozacki, Product Manager on GKE for AI Training, and Wojciech Tyczyński, Software Engineer on the GKE team at Google. We explore what it means for GKE to support 65k nodes, and the open source contributions that made this possible
Do you have something cool to share? Some questions? Let us know:
- web: kubernetespodcast.com
- mail: kubernetespodcast@google.com
- twitter: @kubernetespod
News of the week The Kubernetes Podcast is on Bluesky
OpenTelemetry expanding into CI/CD observability
Gitpod is moving away from Kubernetes
OpenCost is a CNCF Incubated project
Links from the interview Guests:
Maciek
Wojciech
Kubernetes OSS Scalability thresholds
PGS on the Kubernetes Podcast
Batch Working Group
Serving Working Group episode on the podcast
Dynamic Resource Allocation
Kueue
Multitenancy and Fairness at Scale with Kueue
SIG Scalability
Links from the post-interview chat Consistent Reads from Cache
Kubernetes Scalability: A Multi-Dimensional Analysis
Scalability, with Wojciech Tyczynski
Before Kubernetes was launched, it could have at most 25 nodes in a cluster. At 1.0, the target was 100. Meanwhile, Borg, Omega and Mesos were all running away at 10,000. What did it take to get Kubernetes to this number, and above? SIG Scalability and GKE Tech Lead Wojciech Tyczynski tells us.
Do you have something cool to share? Some questions? Let us know:
web: kubernetespodcast.com
mail: kubernetespodcast@google.com
twitter: @kubernetespod
Chatter of the week Follow-up: Chairs, from Episode 107
Christmas trees, from Episode 104
Kids music The duck song
The duck joke
Autotune the News
The duck song goes viral on TikTok
Walmart Yodeling Kid
News of the week KubeCon US goes virtual
PromCon schedule
AWS App2Container Episode 48, with Issy Ben-Shaul
GKE brings Node Local DNS cache to GA Episode 106, with John Belamaric
Update kernel and Kubelet config on GKE nodes
AKS brings 1.17 to GA; adds containerd and priority placement group support
Diamanti Spektra 3.0
Kubernetes WG Naming
Introducing Cloud Native Community Groups
Updated CNCF Storage whitepaper
Presslabs moves to Kubernetes Presslabs Stack and WordPress Operator
Links from the interview Omega Episode 43, with Brian Grant
Defining scalability
Original SLOs API-responsiveness: 99% of all our API calls return in less than 1 second
Pod startup time: 99% of pods (with pre-pulled images) start within 5 seconds
Target SLO doc - 25 nodes
Borg - ~10,000 nodes
Sep 2015, Kubernetes 1.0 - 100 nodes "Kubernetes Has A Ways To Go To Scale Like Google, Mesos" by Timothy Prickett Morgan
March 2016, Kubernetes 1.2 - 1,000 nodes
July 2016, Kubernetes 1.3 - 2,000 nodes Work by Clayton Coleman, guest of Episode 85
March 2017, Kubernetes 1.6 - 5000 nodes
etcd v3 improvements for web scale
Scalability Envelope
Today's scalability numbers
EndpointSlices Episode 104, with Bowei Du
JD.com's 10,000 node clusters
Alibaba's 10,000 node clusters Episode 95, with Xiang Li
Google's 15,000 node GKE clusters
Twitter session at the upcoming Google Cloud Next by Reza Motamedi and Maciek Różacki
Poseidon and Firmament
Wojciech Tyczynski: GitHub
LinkedIn