ASPLOS 2021 Artifacts
These are the artifacts for the ASPLOS 2021 paper: Rethinking Software Runtimes for Disaggregated Memory.
Instructions
These instructions have been tested on a clean Ubuntu 20.04 installation running on a CloudLab C6420 machine.
Make sure you run everything on a bare-metal Linux server, that you have sudo access and at least 128 GB RAM and 100 GB free storage space for application datasets and logs.
It is best to launch everything inside a screen session.
Clone the repository and submodules
git clone --recurse-submodules https://github.com/project-kona/asplos21-ae.git
cd asplos21-ae
Applications
Set up applications and download data sets. To do so, run the following from the asplos21-ae folder:
cd apps/scripts
./setup.sh
KTracker
Run all applications and produce the KTracker data. To do so, run the following from the asplos21-ae folder:
cd KTracker/scripts
./run_all.sh
All output data and plots will be generated in the results directory.
For each run, a new directory is created res_<date>.
The generated plots are in a plots directory, under this res_<date> directory (redis_amplif.pdf and results_wp.pdf).
KCacheSim
Refer to KCacheSim repository for more details.
Navigate to KCacheSim directory
cd KCacheSim
Install dependencies
./scripts/setup.sh
Run everything (this will take a long time and it is best to launch this inside a screen session)
python3 ./scripts/sweep.py
All logs will be generated in logs directory
Finally, generate all plots
python3 ./scripts/gather-results.py
All plots will be generated in plots directory.