DART

[SIGMOD'26]: DART: A Lock-free Two-layer Hashed ART Index for Disaggregated Memory

Compile

Clone

git clone <this repo>
git submodule update --init --recursive

Build

sudo apt install libboost-context-dev libboost-coroutine-dev
cmake -B build
cmake --build build

Workload Generation

./script/workload_download.py
./script/workload_gen.sh

./script/split_and_send_workload.py --inputs a_load a_run --outputs a_load_split a_run_split --ips 70 72 74

Run

sudo sysctl -w vm.nr_hugepages=16384

bin/monitor --test_func=0 --memory_num=1 --compute_num=1 --load_thread_num=56 --run_thread_num=56 --coro_num=1 --mem_mb=8192 --th_mb=10 --workload_load=c_load --workload_run=c_run --bucket=256 --run_max_request=6000000
bin/memory --monitor_addr=127.0.0.1:9898  
bin/compute --monitor_addr=127.0.0.1:9898  

Note

The RACE hashing part of this code is based on the implementation from: https://github.com/minxinhao/SepHash

You are welcome to substitute this component with a more efficient hash table if one is available.