Skip to content

Exploiting Cloud Services for Cost-Effective, SLO-Aware Machine Learning Inference Serving

Notifications You must be signed in to change notification settings

MachineLearningSystem/MArk-Project

This branch is up to date with marcoszh/MArk-Project:master.

Folders and files

NameName
Last commit message
Last commit date

Latest commit

180be0a · Dec 27, 2019

History

3 Commits
Jan 8, 2019
Jan 8, 2019
Oct 27, 2019
Jan 8, 2019
Jan 8, 2019
Jan 8, 2019
Jan 8, 2019
Jan 8, 2019
Jan 8, 2019
Dec 27, 2019
Jan 8, 2019

Repository files navigation

MArk

Instructions

You need to build your own AWS AMI that encapsulates the actual serving backend first to be able to run the experiments, the AMI must be able to be requested following the patterns in model_source.py. The easiest way to do so would be using a MXNet Model Server image.

pre-launch instances

  • modify the instance type and model type in constants.py (INS_SOURCE, MODEL)
  • cmd : ./bin/start_server.sh launch $tag(optional,default 0)

run experiment

  • run frontend: ./bin/start_server.sh start $tag(optional,default 0)
  • modify which sender to use in experiment/request_sender.py
  • run request sending process: ./bin/start_server.sh send $burst(optional)

collect log

  • move log to assigned dir: ./bin/start_server.sh move $tag(prefix name of log dir, e.g. $tag-v1)
  • parse latency from log: python3 experiment/parser/parse_latency.py $path_to_log_dir

About

Exploiting Cloud Services for Cost-Effective, SLO-Aware Machine Learning Inference Serving

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 69.6%
  • C++ 29.8%
  • C 0.6%
  • Fortran 0.0%
  • CSS 0.0%
  • JavaScript 0.0%