Skip to content

Latest commit

 

History

History
77 lines (55 loc) · 2.32 KB

README.md

File metadata and controls

77 lines (55 loc) · 2.32 KB
Raw

LHR Simulator

Here provides the basic realization of LHR on Python3. The simulator here contains all functions described in our submitted paper "Learning from Optimal Caching for Content Delivery".

Usage

  1. Requirement: Python >= 3.6, Xgboost and Lightgbm

  2. The script "LHR.py" is the realization of LHR algorithm, parameters can be changed under different cases

  3. The code named "Utils.py" contains the functions used in LHR

  4. Run the file "Main.py" to start the LHR

  5. "HRO.py" provides the method to get the HRO upper bound and "Belady.py" simulates the behaviour of Belady-Size algorithm

Trace Format

Request traces are expected to be in a space-separated format with 3 columns: request time, content id and content size.

Time ID Size
0 1 1024
1 23 1024
2 15 2048
3 10 1024

Experiments

We have implemented LHR and a bunch of state-of-the-arts in simulators including but not limited to:

  • LHR
  • Hawkeye
  • Learning Relaxed Belady (LRB)
  • LRU
  • B-LRU (Bloom Filter LRU)
  • LRUK
  • LFUDA
  • AdaptSize

To get the results in figure 8 of our paper, you should do:

  • Run LHR algorithm to get the real hits
  • Run Simulators in Webcachesim to get results for SOTAs
  • By using real-hit results, run the function in "Utils.py" to get the estimated traffic

To get figure 2, you should do:

  • Run "HRO.py" and "Belady.py" to get HRO bound and Belady bound
  • Run PFOO to get the PFOO bound
  • SOTAs could be reached by using Webcachesim

To estimate overhead, put all algorithms on the same platform and add the following codes to estimate the overhead:

import time
TStart = time.time()
......
TEnd = time.time()
RunTime = TEnd - TStart

import psutil
CPU = psutil.cpu_percent()
Memory = psutil.virtual_memory()

Citation

If you use the simulator or some results in our paper for a published project, please cite our work by using the following bibtex entry

@inproceedings{yan2021learning,
  title={Learning from Optimal Caching for Content Delivery},
  author={Yan, Gang and Li, Jian and Towsley, Don},
  booktitle={Proc. of ACM CoNEXT},
  year={2021}
}