Deft: A Scalable Tree Index for Disaggregated Memory

This repository contains the implementaion of Deft. Deft is a tree-based index for disaggregated memory that delivers high performance and scalability.

For more details, please refer to our paper: [EuroSys'25]: Deft: A Scalable Tree Index for Disaggregated Memory.

This work is built on top of our previous work Sherman.

System Requirements

1. Mellanox ConnectX-5 NICs and above
2. RDMA Driver: MLNX_OFED_LINUX-4.9-5.1.0.0 (If you use MLNX_OFED_LINUX-5**, you should modify codes to resolve interface incompatibility)
3. NIC Firmware: version 16.26.4012 and above (to support on-chip memory, you can use ibstat to obtain the version)
4. memcached (to exchange QP information)
5. cityhash
6. boost 1.53 and above (to support boost::coroutines::symmetric_coroutine)

Setup about RDMA Network

1. RDMA NIC Selection.

You can modify this line according the RDMA NIC you want to use, where ibv_get_device_name(deviceList[i]) is the name of RNIC (e.g., mlx5_0)

2. Gid Selection.

If you use RoCE, modify gidIndex in this line according to the shell command show_gids, which is usually 3.

3. MTU Selection.

If you use RoCE and the MTU of your NIC is not equal to 4200 (check with ifconfig), modify the value path_mtu in src/rdma/StateTrans.cpp

4. On-Chip Memory Size Selection.

Change the constant kLockChipMemSize in include/Commmon.h, making it <= max size of on-chip memory.

Getting Started

Manually (similar to Sherman)

• cd deft

• ./script/hugepage.sh to request huge pages from OS (use ./script/clear_hugepage.sh to return huge pages)

• mkdir build; cd build; cmake -DCMAKE_BUILD_TYPE=Release ..; make -j

• cp ../script/restartMemc.sh .

• configure ../memcached.conf, where the 1st line is memcached IP, the 2nd is memcached port

• ./restartMemc.sh (to initialize memcached server)

• (You can use one numa node with one RDMA NIC to simulate one server or one client.)

• For each server, execute ./server --server_count kServerCount --client_count kClientCount --numa_id 0

• For each client, execute ./client --server_count kServerCount --client_count kClientCount --numa_id 0

With scripts

• Mount this repository with nfs on all servers and clients with the same path.
• Edit script/global_config.yaml with reference to the template script/global_config_sample.yaml, to specify the server and client information. This config will be used by other scripts.
• mkdir build; cd build; cmake -DCMAKE_BUILD_TYPE=Release ..; make -j
• ../script/all_hugepage.py to request huge pages in all servers.
• ../script/run_bench.py, which will run benchmarks of the product of the paramters.
  • You can edit script/run_bench.py to change the parameters of the benchmark (e.g., threads_CN_arr, read_ratio_arr (type: python list)).
  • logs and results will be saved in the log and result directories, respectively.

Citation

If you use Deft in your research, please cite our paper:

@inproceedings{deft2025eurosys,
    author = {Wang, Jing and Wang, Qing and Zhang, Yuhao and Shu, Jiwu},
    title = {Deft: A Scalable Tree Index for Disaggregated Memory},
    year = {2025},
    isbn = {9798400711961},
    publisher = {Association for Computing Machinery},
    address = {New York, NY, USA},
    url = {https://doi.org/10.1145/3689031.3696062},
    doi = {10.1145/3689031.3696062},
    booktitle = {Proceedings of the Twentieth European Conference on Computer Systems},
    // pages = {TO BE FILLED},
    numpages = {16},
    keywords = {Disaggregated Memory, Index, RDMA},
    location = {Rotterdam, Netherlands},
    series = {EuroSys '25}
}