Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FaasMeter: Energy-First Serverless Computing

Abdul Rehman, Alexander Fuerst, Prateek Sharma

TL;DR

FaasMeter introduces a full-system energy metrology framework for Functions as a Service (FaaS), elevating energy to a first-class resource in serverless platforms. By fusing direct attribution with model-based disaggregation, online Kalman filtering, CPU power modeling, and Shapley-value-inspired fair sharing, it provides per-invocation footprints and fair accounting that support energy capping and pricing. External validation against marginal energy demonstrates high accuracy across diverse hardware, while internal metrics show footprint stability and symmetry suitable for pricing. The approach enables energy-aware scheduling and governance in heterogeneous, multi-tenant serverless environments, with practical impact for decarbonization and cost efficiency.

Abstract

Functions as a Service has emerged as a popular abstraction for a wide range of cloud applications and an important cloud workload. We present the design and implementation of FaasMeter, a FaaS control plane which provides energy monitoring, accounting, control, and pricing as first-class operations. The highly diverse and dynamic workloads of FaaS create additional complexity to measuring and controlling energy usage which FaasMeter can mitigate. We develop a new statistical energy disaggregation approach to provide accurate and complete energy footprints for functions, despite using noisy and coarse-grained system-level power (not just CPU power readings). Our accurate and robust footprints are achieved by combining conventional power models with Kalman filters and Shapley values. FaasMeter is a full-spectrum energy profiler, and fairly attributes energy of shared resources to functions (such as energy used by the control plane itself). We develop new energy profiling validation metrics, and show that FaasMeter's energy footprints are accurate to within 1\% of carefully obtained marginal energy ground truth measurements.

FaasMeter: Energy-First Serverless Computing

TL;DR

FaasMeter introduces a full-system energy metrology framework for Functions as a Service (FaaS), elevating energy to a first-class resource in serverless platforms. By fusing direct attribution with model-based disaggregation, online Kalman filtering, CPU power modeling, and Shapley-value-inspired fair sharing, it provides per-invocation footprints and fair accounting that support energy capping and pricing. External validation against marginal energy demonstrates high accuracy across diverse hardware, while internal metrics show footprint stability and symmetry suitable for pricing. The approach enables energy-aware scheduling and governance in heterogeneous, multi-tenant serverless environments, with practical impact for decarbonization and cost efficiency.

Abstract

Functions as a Service has emerged as a popular abstraction for a wide range of cloud applications and an important cloud workload. We present the design and implementation of FaasMeter, a FaaS control plane which provides energy monitoring, accounting, control, and pricing as first-class operations. The highly diverse and dynamic workloads of FaaS create additional complexity to measuring and controlling energy usage which FaasMeter can mitigate. We develop a new statistical energy disaggregation approach to provide accurate and complete energy footprints for functions, despite using noisy and coarse-grained system-level power (not just CPU power readings). Our accurate and robust footprints are achieved by combining conventional power models with Kalman filters and Shapley values. FaasMeter is a full-spectrum energy profiler, and fairly attributes energy of shared resources to functions (such as energy used by the control plane itself). We develop new energy profiling validation metrics, and show that FaasMeter's energy footprints are accurate to within 1\% of carefully obtained marginal energy ground truth measurements.
Paper Structure (22 sections, 6 equations, 12 figures, 3 tables)

This paper contains 22 sections, 6 equations, 12 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Functions as a Service (FaaS)
  4. Energy-first Cloud Computing
  5. Challenges in Function-level Energy Measurement
  6. Direct Attribution
  7. Model-based Power Estimation
  8. Validation
  9. Shared Components
  10. FaasMeter Energy Profiling
  11. Statistical Power Disaggregation
  12. Online Estimation with Kalman Filtering
  13. CPU Power Modeling
  14. Fairly Attributing Shared Power
  15. Implementation and Validation Methods
  16. ...and 7 more sections

Figures (12)

  • Figure 1: FaasMeter integrates energy-awareness into the FaaS control plane. Multiple power sources are used to obtain accurate per-function energy footprints, which we use for energy capping and pricing.
  • Figure 2: Function power signatures cannot be captured reliably by existing power profiling methods.
  • Figure 3: Function energy per invocation, measured in isolation. Server load and concurrency levels significantly impact the footprints, making this an unreliable method for energy measurement.
  • Figure 4: FaasMeter adapts a Kalman filter approach forupdating function per-invocation power $X$ over time.
  • Figure 5: Synchronization of system-level power.
  • ...and 7 more figures