Clevel Hashing

Introduction

Clevel hashing is a lock-free PM hashing index with asynchronous resizing, which supports lock-free operations for search/insertion/update/deletion. This repository is the implementation of "Lock-free Concurrent Level Hashing for Persistent Memory", in Proceedings of the USENIX Annual Technical Conference (USENIX ATC), 2020.

Implementation & Tests

We implement clevel hashing with PMDK and provide query interfaces similar to the concurrent_hash_map in libpmemobj-cpp. The implementation and tests for clevel hashing are based on the libpmemobj-cpp (commit@26c86b4699).

Compilation

Requirements:

• cmake >= 3.3
• libpmemobj-dev(el) >= 1.4 (http://pmem.io/pmdk/)
• compiler with C++11 support

How to build

We only ran our tests on Linux (Ubuntu 18.04). Run the following commands for compilation:

$ mkdir build
$ cd build
$ cmake ..
$ make -j

To compile with TBB enabled, please run the following commands:

$ mkdir build
$ cd build
$ cmake .. -DUSE_TBB=1
$ make -j

Run

Please refer to the instructions for the use of test files.

Limitation

• Current tests are used for the performance evaluation and not designed for ctest.

Contact

If you have any problems, please feel free to contact me.

• Zhangyu Chen (chenzy@hust.edu.cn)

****Following documentation comes from original libpmemobj-cpp****

libpmemobj-cpp

C++ bindings for libpmemobj (https://github.com/pmem/pmdk) More information in include/libpmemobj++/README.md

How to build

Requirements:

• cmake >= 3.3
• libpmemobj-dev(el) >= 1.4 (http://pmem.io/pmdk/)
• compiler with C++11 support

On Linux

$ mkdir build
$ cd build
$ cmake ..
$ make
$ make install

When developing:

$ ...
$ cmake .. -DCMAKE_BUILD_TYPE=Debug -DDEVELOPER_MODE=1
$ ...
$ ctest --output-on-failure

To build packages

...
cmake .. -DCPACK_GENERATOR="$GEN" -DCMAKE_INSTALL_PREFIX=/usr
make package

$GEN is type of package generator and can be RPM or DEB

CMAKE_INSTALL_PREFIX must be set to a destination were packages will be installed

To use Intel(R) Threading Building Blocks library

By default concurrent_hash_map uses pmem::obj::shared_mutex internally. But read-write mutex from Intel(R) Threading Building Blocks library can be used instead to achieve better performance. To enable it in your application set the following compilation flag:

• -DLIBPMEMOBJ_CPP_USE_TBB_RW_MUTEX=1

If you want to build tests for concurrent_hash_map with read-write mutex from Intel(R) Threading Building Blocks library, run cmake with -DUSE_TBB=1 -DTBB_DIR=<Path to Intel TBB>/cmake option.

Intel(R) Threading Building Blocks library can be downloaded from the official release page.

To use with Valgrind

In order to build your application with libpmemobj-cpp and pmemcheck / memcheck / helgrind / drd, Valgrind instrumentation must be enabled during compilation by adding flags:

• LIBPMEMOBJ_CPP_VG_PMEMCHECK_ENABLED=1 for pmemcheck instrumentation,
• LIBPMEMOBJ_CPP_VG_MEMCHECK_ENABLED=1 for memcheck instrumentation,
• LIBPMEMOBJ_CPP_VG_HELGRIND_ENABLED=1 for helgrind instrumentation,
• LIBPMEMOBJ_CPP_VG_DRD_ENABLED=1 for drd instrumentation, or
• LIBPMEMOBJ_CPP_VG_ENABLED=1 for all Valgrind instrumentations (including pmemcheck).

If there are no memcheck / helgrind / drd / pmemcheck headers installed on your system, build will fail.

On Windows

Install libpmemobj via vcpkg

vcpkg install pmdk:x64-windows
vcpkg integrate install

...
cmake . -Bbuild -G "Visual Studio 14 2015 Win64"
        -DCMAKE_TOOLCHAIN_FILE=c:/tools/vcpkg/scripts/buildsystems/vcpkg.cmake

msbuild build/ALL_BUILD.vcxproj