Red Hat
Senior Principal Software Engineer
Currently a Senior Principal Software Engineer at Red Hat, Sunil Thaha is passionate about sustainable computing and power efficiency in cloud-native environments, architecting solutions that reduce resource consumption while maintaining enterprise-grade performance. He has long been an active contributor to CNCF projects, focusing on the development and advancement of Kubernetes ecosystem tools. As Project Lead of Kepler, he drives technical innovation in power monitoring for Kubernetes workloads. He has extensive experience leading technical discussions across global open-source communities and mentoring contributors, dedicated to advancing cloud-native technologies and sustainable computing practices.
(AI 翻譯)
Sunil Thaha 目前擔任 Red Hat 的 資深首席軟體工程師,專注於雲原生環境中的永續運算與電力效率,致力於設計在維持企業級效能的同時,能降低資源消耗的解決方案。
他長期活躍於 CNCF 專案貢獻,特別聚焦於 Kubernetes 生態系工具的發展與推進。作為 Kepler 專案負責人,他推動 Kubernetes 工作負載的電力監控技術創新。
Sunil 也具備豐富的經驗,領導全球開源社群的技術討論並指導貢獻者,持續推動雲原生技術與永續運算實踐的發展。
(AI 翻譯)
隨著 Kubernetes 叢集在全球規模擴展,能源消耗已成為關鍵的營運課題,但多數組織對雲原生工作負載的實際用電量卻缺乏可見性。本演講將介紹 Kepler (Kubernetes-based Efficient Power Level Exporter),這是一個 CNCF 沙盒專案,能將能源消耗從「黑盒子」轉化為可操作的 Prometheus 指標。
Kepler 透過整合 Intel RAPL 感測器 與 CPU 使用模式,在容器、Pod、VM 與節點層級提供細緻的能耗歸因。不同於傳統基礎架構監控僅著眼於資源需求與限制,Kepler 能揭露工作負載的實際能源成本,讓使用者能以數據驅動決策,推動永續運算。
我們將深入探討 Kepler 全新改寫的 v0.10.x 架構,該版本效能大幅提升,且僅需對 /proc 與 /sys 具備讀取權限(無需再使用特權能力)。本場次也將示範手動部署與 Kepler Operator,後者透過 PowerMonitor 自訂資源(CRD) 提供宣告式設定,實現自動化部署與生命週期管理。
與會者將學到如何:
透過現場示範,我們將展示多種情境,包括容器、Pod、VM 工作負載的能耗歸因,以及透過 Redfish BMC 整合的平臺層級電力監控。參加者將帶回實用工具,能夠量測、監控並最佳化 Kubernetes 基礎架構的環境影響。
無論你正在管理受限於電力條件的邊緣部署,或是致力於達成永續目標的大型資料中心,本場次都將為你提供打造「節能意識 Kubernetes 營運」的基礎。
目標聽眾:平台工程師、SRE、永續實踐者,以及任何對量測與最佳化雲原生工作負載環境影響有興趣的人。
(Abstract)
As Kubernetes clusters scale globally, energy consumption has become a critical operational concern, yet most organizations lack visibility into the actual power usage of their cloud-native workloads. This talk introduces Kepler (Kubernetes-based Efficient Power Level Exporter), a CNCF sandbox project that transforms energy consumption from a black box into actionable Prometheus metrics.
Kepler provides granular energy attribution at container, pod, VM, and node levels by leveraging Intel RAPL sensors and CPU utilization patterns. Unlike traditional infrastructure monitoring that focuses on resource requests and limits, Kepler reveals the energy cost of your workloads, enabling data-driven decisions for sustainable computing.
We'll explore Kepler's recently rewritten v0.10.x architecture, which delivers enhanced performance while requiring only read access to `/proc` and `/sys` (no more privileged capabilities). The session will demonstrate both manual deployment and the Kepler Operator, which automates deployment and lifecycle management using the PowerMonitor custom resource for declarative configuration.
Attendees will learn how to:
Through live demos, we'll demonstrate scenarios including energy attribution for container, pod, VM workload, and platform-level power monitoring via Redfish BMC integration. You'll leave with practical tools to measure, monitor, and optimize the environmental impact of your Kubernetes infrastructure.
Whether you're managing edge deployments with power constraints or large-scale data centers with sustainability goals, this session provides the foundation for energy-conscious Kubernetes operations.
Target Audience: Platform engineers, SREs, sustainability practitioners, and anyone interested in measuring and optimizing the environmental impact of cloud-native workloads.