在特定平台上，开普勒准确吗？ | KC23

Is Kepler Accurate on Specific Platforms? - Jie Ren & Ken Lu, Intel

在特定平台上，开普勒准确吗？| Is Kepler Accurate on Specific Platforms? - Jie Ren & Ken Lu, Intel

Kepler（Kubernetes高效功率级别导出器）是一个CNCF沙盒项目，它使用eBPF来探测与云原生容器相关的能源统计信息，并将其导出为Prometheus指标。它支持功率比建模和功率估计建模。目前，Kepler的测试用例，无论是单元测试还是集成测试，都是白盒导向的，重点是度量收集和导出代码逻辑检查。在本次演讲中，我们将介绍一个平台验证框架，为Kepler的测试框架提供有益的补充，使开发人员和最终用户能够在特定硬件平台上验证项目功能。供应商特定的验证案例可以集成到框架中。我们可以检查它们的测试报告，以查看供应商的平台在Kepler中是否得到良好支持，评估和识别平台差距和限制，特别是功率建模的准确性。该框架是完全自动化的，具有平台无关性。

Kepler (Kubernetes Efficient Power Level Exporter) is a CNCF Sandbox project, it uses eBPF to probe energy-related system statistics for cloud native containers and exports them as Prometheus metrics. It supports Power Ratio Modeling and Power Estimation Modeling. Currently Kepler’s test cases, both in unit test and integration test, are white-box oriented and focusing on the metrics collection and exportation code logic check. In this talk, we will introduce a platform validation framework to make beneficial supplement to Kepler’s test framework, enable the developers and end users to validate the project features on specific hardware platforms. Vendor specific validation cases could be integrated into the framework. We could check test reports of them to see if the vendors’ platforms are well supported in Kepler, to evaluate and identify the platform gap and limitation, especially on the accuracy of the power modeling. The framework is fully automated and runner system agnostic.

CNCF概况（幻灯片）

扫描二维码联系我们！

CNCF (Cloud Native Computing Foundation)成立于2015年12月，隶属于Linux  Foundation，是非营利性组织。 

CNCF（云原生计算基金会）致力于培育和维护一个厂商中立的开源生态系统，来推广云原生技术。我们通过将最前沿的模式民主化，让这些创新为大众所用。请关注CNCF微信公众号。