Runtime Bad Block Salvation

• Base simulator: SimpleSSD version 2.0

Requirements

• cmake
• g++
• gnu-parallel
• python3

Build

Run ./build.sh && make

Execution

Basic execution: ./simplessd-standalone {config_path} {ssd_config_path} {log_folder_path}

Batch script: ./run.sh

Config

SimpleSSD-Standalone Config

Location: config/(alibaba|exchange|msnfs|systor17)_sampled.cfg

Block Traces

Name	Description
Format	{time} 0 {slba} {nlb} {op}
time	relative time (ns)
slba	start logical block address (512-byte aligned)
nlb	number of logical blocks
op	R|W

Sampled Traces

Write size in trace: 24GB (actual write size must be bigger than OP size)

Trace Location

/mnt/hdd_4/hdd_2/sixtyfive in cluster

Miscellaneous Scripts

Script	Description
stats.cpp	Prints {name}, {# of lines}, {total size}, {total write size}, {bandwidth} in csv format.
sample.py	Parses traces in a given write size from the middle of the total block trace.

Analysis

After baseline and our scheme are evaluated on ./results/,

• tools/analyze.py: converts trace.txt to json format.
• tools/normalize.py: reads json files converted from analyze.py and prints normalized performance in csv format.