SSD performance improvement through deep learning-based workload analysis

2023 Pusan National University Graduation Project

Contributors:

최성찬 (201824641)

Odgerel Ariunbold (201824623)

Ganchuluun Narantsatsralt (201824621)

data folder contains the data to train and currently it is only the sample file.

data/traceProcess is a script to process whole data from ssdtrace-00 to ssdtrace-26 at once

data/ssdtrace.gz is a gz file which is output of above traceProcess script. This will be used for training model.

data/labeling folder contains the labeled data and codes to lable the datas.

data/labeling/Sector.cpp is a code to calculate frequency, access interval, and overall i/o size.

src/ only source codes must be included in the subfolder of src

src/greedyGC Garbage Collection algorithm implementation src

src/model Deep learning models

src/model/cpp Deployment of LSTM Model to C++ environment.

src/utils other useful scripts