Research data supporting "Soroban: Attributing Latency in Virtualized Environments"
Change log
Authors
Description
This is data required for reproducing the figures in the "Soroban: Attributing Latency in Virtualized Environments" paper. It consists of two archives:
[rscfl-exp.zip] containing (i) raw data from kernel level measurements of lighttpd serving requests while running in a virtualized environment (in json format, the .dat and .mdat metadata files) (ii) post-processed measurements with the data filtered and added into python pandas.DataFrame objects for querying and plotting (in standard binary python pickle serialization format, .sdat files) (iii) attribution model files containing serializations of python sklearn.gaussian_process.GaussianProcess objects after training (in standard binary python pickle serialization format, .pickle files) (iv) python scripts for processing and plotting the data (.py files)
AND [rscfl-repr.zip] containing (i) experiment definition files for figures in the paper, containing the information required to reproduce them (.json format) (ii) python script for reproducing experiments (rscfl_exp.py)
Additional information: As an example, for reproducing Figure 3 from the paper, one needs to unpack the two archives (rscfl-repr and rscfl-exp) and run
$ ./rscfl_exp.py -c fig3.config.json -s /path/to/rscfl_exp --fg_load_vm localhost