Implement Speedb Benchmark

.gitignore

@@ -8,3 +8,9 @@ tags
 compile_commands.json
 .clangd/
 .cache/
+
+CMakeFiles/
+CMakeCache.txt
+*.cmake
+
+build/

.gitmodules

@@ -1,3 +1,12 @@
 [submodule "HdrHistogram_c"]
 	path = HdrHistogram_c
 	url = https://github.com/HdrHistogram/HdrHistogram_c.git
+[submodule "unqlite/unqlite"]
+	path = unqlite/unqlite
+	url = https://github.com/symisc/unqlite
+[submodule "terarkdb/terarkdb"]
+	path = terarkdb/terarkdb
+	url = https://github.com/bytedance/terarkdb.git
+[submodule "speedb/speedb"]
+	path = speedb/speedb
+	url = https://github.com/speedb-io/speedb

CMakeLists.txt

@@ -4,6 +4,7 @@
 cmake_minimum_required(VERSION 3.10)
 project(ycsb-cpp LANGUAGES C CXX)
 set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
 set(CMAKE_BUILD_TYPE Release)
 set(CompilerFlags
@@ -25,6 +26,8 @@ option(BIND_ROCKSDB "build with rocksdb" OFF)
 option(BIND_LMDB "build with lmdb" OFF)
 option(BIND_LEVELDB "build with leveldb" OFF)
 option(BIND_WIREDTIGER "build with wiredtiger" OFF)
+option(BIND_UNQLITE "build with unqlite" OFF)
+option(BIND_TERARKDB "build with terarkdb" OFF)
 
 option(WITH_ZLIB "linking YCSB with zlib; needed by HdrHISTOGRAM, DO NOT TURN OFF" ON)
 option(WITH_LZ4 "linking YCSB with lz4" OFF)
@@ -48,7 +51,7 @@ if (BIND_ROCKSDB)
 
     find_package(RocksDB CONFIG)
     if(RocksDB_FOUND)
-        target_link_libraries(ycsb PRIVATE RocksDB::rocksdb rpcrt4.lib)
+        target_link_libraries(ycsb PRIVATE RocksDB::rocksdb)
     else()
         message(STATUS "Try to find rocksdb library using find_library")
         find_library(ROCKSDB_LIB rocksdb REQUIRED)
@@ -194,8 +197,47 @@ else()
     message(STATUS "WITH_BZ2 - OFF")
 endif()
 
+if (BIND_UNQLITE)
+    message(STATUS "BIND_UNQLITE - ON")
+    file(GLOB_RECURSE YCSB_UNQLITE_SRC "unqlite/*.cc")
+    target_sources(ycsb PRIVATE ${YCSB_UNQLITE_SRC})
+    file(GLOB_RECURSE YCSB_UNQLITE_LIB "unqlite/unqlite/unqlite.c")
+    target_sources(ycsb PRIVATE ${YCSB_UNQLITE_LIB})
+else()
+    message(STATUS "BIND_UNQLITE - OFF")
+endif()
+
+if (BIND_TERARKDB)
+    message(STATUS "BIND_TERARKDB - ON")
+
+    file(GLOB YCSB_TERARKDB_SRC "terarkdb/*.cc")
+    target_sources(ycsb PRIVATE ${YCSB_TERARKDB_SRC})
+
+    target_link_libraries(ycsb PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/terarkdb/terarkdb/build/build/lib/libterarkdb.a")
+    target_include_directories(ycsb PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/terarkdb/terarkdb/build/build/include")
+
+    # link all of terarkdb's dependencies
+    set(TERARKDB_BUILD_DIR "${CMAKE_CURRENT_SOURCE_DIR}/terarkdb/terarkdb/build")
+    set(TERARKDB_LIB_DEPS_DIR "${TERARKDB_BUILD_DIR}/lib")
+    find_library(BZ2_LIB bz2 HINTS ${TERARKDB_LIB_DEPS_DIR} REQUIRED)
+    target_link_libraries(ycsb PRIVATE ${BZ2_LIB})
+    find_library(LZ4_LIB lz4 HINTS ${TERARKDB_LIB_DEPS_DIR} REQUIRED)
+    target_link_libraries(ycsb PRIVATE ${LZ4_LIB})
+    find_library(ZSTD_LIB zstd HINTS ${TERARKDB_LIB_DEPS_DIR} REQUIRED)
+    target_link_libraries(ycsb PRIVATE ${ZSTD_LIB})
+    find_library(JEMALLOC_LIB jemalloc HINTS ${TERARKDB_LIB_DEPS_DIR} REQUIRED)
+    target_link_libraries(ycsb PRIVATE ${JEMALLOC_LIB})
+    find_library(SNAPPY_LIB snappy HINTS ${TERARKDB_LIB_DEPS_DIR} REQUIRED)
+    target_link_libraries(ycsb PRIVATE ${SNAPPY_LIB})
+    find_library(Z_LIB z HINTS ${TERARKDB_LIB_DEPS_DIR} REQUIRED)
+    target_link_libraries(ycsb PRIVATE ${Z_LIB})
+    target_include_directories(ycsb PRIVATE "${CMAKE_CURRENT_SOURCE_DIR}/terarkdb/terarkdb/build/include")
+else()
+    message(STATUS "BIND_TERARKDB - OFF")
+endif()
+
 add_subdirectory(HdrHistogram_c)
 include_directories(HdrHistogram_c/include)
 add_compile_definitions(HDRMEASUREMENT)
 add_dependencies(ycsb hdr_histogram_static)
-target_link_libraries(ycsb PRIVATE hdr_histogram_static)
+target_link_libraries(ycsb PRIVATE hdr_histogram_static)

PROJECT_README.md

@@ -0,0 +1,56 @@
+# Exposing the “Bleeding-Edge” of Modern Embedded Key-Value Databases
+
+- **CS 6400 DB Project (Group 8)**
+- _Members: Richard So, Nandha Sundaravadivel, Drumil Deliwala, Saatvik Agrawal_
+
+## Results
+
+## Setup / Instructions
+
+### Prerequisites
+
+- AWS EC2 Instance, `t2.micro` (although any instance should work)
+
+  - Amazon Linux 2 AMI (or any Linux distro)
+
+- Install the following:
+
+  - `cmake`, `make`, `gcc`, `g++`
+  - Install `zlib-devel` or equivalent for compiling `YCSB-cpp`
+  - `java` (8+) and `maven` (>= 3); using `sdkman` is recommended
+  - `python3` for running `YCSB`
+
+### Running YCSB (Java) benchmarks
+
+> **Implemented bindings: `xodus`, `mapdb`, `halodb`**
+
+Below is an example of running the `mapdb` binding:
+
+```sh
+cd YCSB
+./bin/ycsb.sh load mapdb -s -P workloads/workloada
+./bin/ycsb.sh run mapdb -s -P workloads/workloada
+```
+
+Rinse and repeat for the other bindings.
+
+### Running YCSB-cpp (C++) benchmarks
+
+> **Impmeneted bindings: `unqlite`, `terarkdb`**
+
+Below is an example of compiling and running the `unqlite` binding:
+
+```sh
+cd YCSB-cpp
+mkdir build && cd build
+cmake -DBIND_UNQLITE=1 ..
+make clean && make
+./ycsb -load -db unqlite -P ../workloads/workloada -s
+./ycsb -run -db unqlite -P ../workloads/workloada -s
+```
+
+There are additional instructions in the `YCSB-cpp` directory for `terarkdb`,
+`leveldb`, and `rocksdb`.
+
+**It is recommended to only bind ONE database at a time when compiling w/
+cmake to avoid conflicts.**

README.md

@@ -7,13 +7,80 @@ This is a fork of [YCSB-C](https://github.com/basicthinker/YCSB-C) with some add
  * Modified the workload more similar to the original YCSB
  * Supported databases: LevelDB, RocksDB, LMDB, WiredTiger, SQLite
 
+> **Additionally, this repository is an extension of the original
+[YCSB-cpp](https://github.com/ls4154/YCSB-cpp) project, now supporting a few
+more databases (e.g. Speedb, TerarkDB, UnQLite, etc.) and including instructions
+to build and install DBs from source (locally, without root).**
+>
+> _**This specific fork serves as a portion of an end-of-semester project for
+Georgia Tech's CS 6400 Databases course**_
+
 # Build YCSB-cpp
 
+## Prerequisites
+
+  * Clone this repository, recursing to all submodules
+  ```sh
+  git clone --recurse-submodules git@github.com:richardso21/YCSB-cpp.git
+  ```
+
+  * Ensure you have `gcc` and `cmake` installed
+
+### Installing LevelDB
+
+  * LevelDB is included as a submodule in this repository. To build it, run the following commands:
+  ```sh
+  cd YCSB-cpp/third-party/leveldb
+  mkdir build && cd build
+
+  # to install globally...
+  cmake -DCMAKE_BUILD_TYPE=Release ..
+
+  # or, to install locally (e.g. if you are on PACE)
+  cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=~/.local ..
+
+  make install
+  ```
+
+  * **If you installed locally**, you will need to add the following to your
+  `.bashrc` or `.bash_profile` (or `zsh` equivalents):
+  ```bash
+  export PATH=$PATH:~/.local/bin
+  export CPATH=$CPATH:~/.local/include
+  export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:~/.local/lib
+  ```
+> _LevelDB is included in order to provide a reference point for the original
+`YCSB` implementation in Java._
+
+### Building TerarkDB
+
+  * TerarkDB is included as a submodule, found within `YCSB-cpp/terarkdb/terarkdb`.
+
+  * **IMPORTANT**: If you are on PACE, or you do not have XSL (or you do not know what it is),
+  you will need to modify `terarkdb/terarkdb/third-party/CMakeLists.txt` at line 203. Change
+  the line from:
+  ```cmake
+  INSTALL_COMMAND $(MAKE) install PREFIX=${CMAKE_BINARY_DIR}
+  ```
+  to:
+  ```cmake
+  INSTALL_COMMAND $(MAKE) install_bin install_include install_lib PREFIX=${CMAKE_BINARY_DIR}
+  ```
+
+  * To build TerarkDB as a static library for YCSB-cpp, run the following commands:
+  ```sh
+  cd YCSB-cpp/terarkdb/terarkdb
+  WITH_TESTS=OFF WITH_ZNS=OFF ./build.sh
+  ```
+
+> **NOTE**: TerarkDB is a fork of RocksDB, and thus shares the same header paths.
+  **Do not compile YCSB-cpp with both TerarkDB and RocksDB binded simultaneously!**
+
 ## Build with Makefile on POSIX
 
 Initialize submodule and use `make` to build.
 
-```
+```sh
 git clone https://github.com/ls4154/YCSB-cpp.git
 cd YCSB-cpp
 git submodule update --init
@@ -23,20 +90,20 @@ make
 Databases to bind must be specified as build options. You may also need to add additional link flags (e.g., `-lsnappy`).
 
 To bind LevelDB:
-```
+```sh
 make BIND_LEVELDB=1
 ```
 
 To build with additional libraries and include directories:
-```
+```sh
 make BIND_LEVELDB=1 EXTRA_CXXFLAGS=-I/example/leveldb/include \
                     EXTRA_LDFLAGS="-L/example/leveldb/build -lsnappy"
 ```
 
 Or modify config section in `Makefile`.
 
 RocksDB build example:
-```
+```sh
 EXTRA_CXXFLAGS ?= -I/example/rocksdb/include
 EXTRA_LDFLAGS ?= -L/example/rocksdb -ldl -lz -lsnappy -lzstd -lbz2 -llz4
 
@@ -45,7 +112,7 @@ BIND_ROCKSDB ?= 1
 
 ## Build with CMake on POSIX
 
-```shell
+```sh
 git submodule update --init
 mkdir build
 cd build
@@ -60,22 +127,22 @@ see [BUILD_ON_WINDOWS](BUILD_ON_WINDOWS.md)
 ## Running
 
 Load data with leveldb:
-```
+```sh
 ./ycsb -load -db leveldb -P workloads/workloada -P leveldb/leveldb.properties -s
 ```
 
 Run workload A with leveldb:
-```
+```sh
 ./ycsb -run -db leveldb -P workloads/workloada -P leveldb/leveldb.properties -s
 ```
 
 Load and run workload B with rocksdb:
-```
+```sh
 ./ycsb -load -run -db rocksdb -P workloads/workloadb -P rocksdb/rocksdb.properties -s
 ```
 
 Pass additional properties:
-```
+```sh
 ./ycsb -load -db leveldb -P workloads/workloadb -P rocksdb/rocksdb.properties \
     -p threadcount=4 -p recordcount=10000000 -p leveldb.cache_size=134217728 -s
 ```