eflect

eflect is an energy profiler for multi-threaded Java applications in shared environments. Below is our paper abstract about this work:

Developing energy-aware applications is a well known approach to software-based energy optimization. This promising approach is however faced with a significant hurdle when deployed to the environments shared among multiple applications, where the energy consumption effected by one application may erroneously be observed by another application. We introduce eflect, a novel software framework for disentangling the energy consumption of co-running applications. Our key idea, called energy virtualization, enables each energy-aware application to be only aware of the energy consumption effected by its execution. eflect is unique in its lightweight design: it is a purely application-level solution that requires no modification to the underlying hardware or system software. Experiments show eflect incurs low overhead with high precision. Furthermore, it can seamlessly port existing application-level energy frameworks --- one for energy-adaptive approximation and the other for energy profiling --- to shared environments while retaining their intended effectiveness.

New

Building and execution

cd eflect/
nix-shell
gradle build
cd experiments/
sudo ./aeneas-experiments.sh data 500

Development machine

The configuration of the development machine is here. The msr kernel module is loaded automatically.

Old

Experiment reproduction: Docker-based execution

Our publication data can be reproduced using a docker image. You can either run the image from docker hub:

docker run --privileged --cap-add=ALL -it -v /dev:/dev -v /lib/modules:/lib/modules pleflect/eflect:latest

or build and run the provided Dockerfile:

docker build -t eflect-icse22 .
docker run --privileged --cap-add=ALL -it -v /dev:/dev -v /lib/modules:/lib/modules pleflect/eflect:latest

This repository's code will already be fully built and the experiments can be run using eflect/experiments/run-experiments.sh, which will output to eflect/experiments/data:

cd eflect/experiments && run-experiments.sh && python3 scripts/processing data

NOTE: The data reported in the paper was produced through an evaluation with the system described below. As energy consumption varies from system to system, e.g., the number of cores, the OS schedulers, the JVM runtime behavior, etc., a reproduction on a different system may not produce identical results as we reported in the paper. Specifically, eflect requires the use of RAPL, which only works on Intel cpus. If you encounter a pread error while running, you may need to run modprobe msr.

• Dual socket Intel E5-2630 v4 2.20 GHz (20 cores)
• Hyper threading enabled
• 64 GB DDR4 RAM
• Debian 4.9 (linux kernel 4.9)
• Debian default powersave governor
• Java 11 Hotspot VM build 11.0.2+9-LTS

Experiment reproduction: Building from source

If you prefer to build eflect from source, please follow the instructions below:

Building

eflect requires the following to build and run:

apt-get install -y git openjdk-11-jdk openjdk-11-dbg libjna-jni git make wget kmod python3 python3-pip
pip3 install numpy pandas matplotlib

A deployable eflect jar can be built from the top-level with make eflect. To run the experiments, you will need to first run eflect/experiments/setup.sh which will download the experiment dependencies and build the driving code.

Execution

Once eflect and its driver are built, the entire experiment can be run using eflect/experiments/run-experiments.sh in the same way as the docker image.