Merge branch 'topling:memtable_as_log_index' into memtable_as_log_index

Author: syslucas

README-zh_cn.md

@@ -1,13 +1,31 @@
 ## ToplingDB: 一个外存上的持久化 Key-Value 存储引擎
 ToplingDB 由[北京拓扑岭科技有限公司](https://topling.cn)开发与维护，从 [RocksDB](https://github.com/facebook/rocksdb) 分叉而来，详情参考 [ToplingDB 分支名称约定](https://github.com/topling/toplingdb/wiki/ToplingDB-Branch-Name-Convention)。
 
+## 快速开始
+ToplingDB 需要 C++17，推荐 gcc 8.3 以上，或者 clang 也行。
+
+ToplingDB 比 RocksDB 快得多，您可以自己快速验证：
+### Compile & run db_bench
+```bash
+sudo yum -y install git libaio-devel gcc-c++ gflags-devel zlib-devel bzip2-devel libcurl-devel liburing-devel snappy-devel jemalloc-devel
+#sudo apt-get update -y && sudo apt-get install -y libjemalloc-dev libaio-dev libgflags-dev zlib1g-dev libbz2-dev libcurl4-gnutls-dev liburing-dev libsnappy-dev libbz2-dev liblz4-dev libzstd-dev
+git clone https://github.com/topling/toplingdb
+cd toplingdb
+make -j`nproc` db_bench DEBUG_LEVEL=0
+sudo make install PREFIX=/some/path # default is /usr/local
+```
+
+以上编译命令执行后，运行 [db_bench.sh](db_bench.sh)(需要[端口 2011](https://github.com/topling/rockside/blob/master/sample-conf/db_bench_enterprise.yaml#L4 "内嵌的 http web 服务使用端口 2011"))，然后使用 ToplingDB：[原生 C++](https://github.com/topling/rockside/wiki/101 "典型场景是从 rocksdb 迁移过来)")，也支持 [Java](https://github.com/topling/rockside/wiki/SidePlugin-Java-Binding "内置在本 github 仓库中") 和 [Rust](https://github.com/topling/rust-toplingdb "另外的专门的 github 仓库")。
+
+## 简单介绍
 ToplingDB 的子模块 **[rockside](https://github.com/topling/rockside)** 是 ToplingDB 的入口，详情参考 **[SidePlugin wiki](https://github.com/topling/rockside/wiki)**。
 
 ToplingDB 兼容 RocksDB API 的同时，增加了很多非常重要的功能与改进：
 1. [SidePlugin](https://github.com/topling/rockside/wiki) 让用户可以通过 json/yaml 文件来定义 DB 配置
 1. [内嵌 Http](https://github.com/topling/rockside/wiki/WebView) 让用户可以通过 Web 查看几乎所有 DB 信息，这是 [SidePlugin](https://github.com/topling/rockside/wiki) 的一个子功能
 1. [内嵌 Http](https://github.com/topling/rockside/wiki/WebView) 让用户可以无需重启进程，[在线修改](https://github.com/topling/rockside/wiki/Online-Change-Options) 各种 db/cf 配置，包括修改 DB 元对象（例如 MemTabFactory, TableFactory, WriteBufferManager ...）
 1. 为提升性能和可扩展性而实施的很多重构与改进，例如 MemTable 的重构
+1. MemTable 可作为 WAL 的索引，消除 MemTable 到 L0 SST 的 Flush，减小写放大，对大尺寸 MemTable 很友好
 1. 对事务处理的改进，特别是 TransactionDB 中 Lock 的管理，热点代码有 5x 以上的性能提升
 1. MultiGet 中使用 fiber/coroutine + io_uring 实现了并发 IO，比 RocksDB 自身的异步 MultiGet 又快又简洁，相应的代码量要少 100 倍不止
 1. [去虚拟化](https://github.com/topling/rockside/wiki/Devirtualization-And-Key-Prefix-Cache-Principle)，消除热点代码中的虚函数调用（主要是 Comparator），并且增加了 Key 前缀缓存，参考相应 [bechmarks](https://github.com/topling/rockside/wiki/Devirtualization-And-Key-Prefix-Cache-Benchmark)
@@ -52,30 +70,6 @@ toplingdb
 
 为了简化编译流程，ToplingDB 在 Makefile 中会自动 clone 各个组件的 github 仓库，社区版用户可以成功 clone 公开的仓库，但克隆私有仓库（例如 topling-rocks）会失败，所以社区版用户编译出来的 ToplingDB 无法创建 Topling**Zip**Table，但可以读取 Topling**Zip**Table。
 
-## 运行 db_bench
-ToplingDB 需要 C++17，推荐 gcc 8.3 以上，或者 clang 也行。
-
-即便没有 Topling**Zip**Table，ToplingDB 也比 RocksDB 要快得多，您可以通过运行 db_bench 来验证性能：
-```bash
-sudo yum -y install git libaio-devel gcc-c++ gflags-devel zlib-devel bzip2-devel libcurl-devel liburing-devel
-#sudo apt-get update -y && sudo apt-get install -y libjemalloc-dev libaio-dev libgflags-dev zlib1g-dev libbz2-dev libcurl4-gnutls-dev liburing-dev libsnappy-dev libbz2-dev liblz4-dev libzstd-dev
-git clone https://github.com/topling/toplingdb
-cd toplingdb
-make -j`nproc` db_bench DEBUG_LEVEL=0
-cp sideplugin/rockside/src/topling/web/{style.css,index.html} ${/path/to/dbdir}
-cp sideplugin/rockside/sample-conf/db_bench_*.yaml .
-export LD_LIBRARY_PATH=`find sideplugin -name lib_shared`
-# change db_bench_community.yaml as your needs
-# 1. use default path(/dev/shm) if you have no fast disk(such as a cloud server)
-# 2. change max_background_compactions to your cpu core num
-# 3. if you have github repo topling-rocks permissions, you can use db_bench_enterprise.yaml
-# 4. use db_bench_community.yaml is faster than upstream RocksDB
-# 5. use db_bench_enterprise.yaml is much faster than db_bench_community.yaml
-# command option -json can accept json and yaml files, here use yaml file for more human readable
-./db_bench -json=db_bench_community.yaml -num=10000000 -disable_wal=true -value_size=20 -benchmarks=fillrandom,readrandom -batch_size=10
-# you can access http://127.0.0.1:2011 to see webview
-# you can see this db_bench is much faster than RocksDB
-```
 ## 可配置的功能
 为了性能和简化，ToplingDB 默认禁用了一些 RocksDB 的功能：
 
@@ -86,8 +80,6 @@ export LD_LIBRARY_PATH=`find sideplugin -name lib_shared`
 宽列 | TOPLINGDB_WITH_WIDE_COLUMNS
 华而不实的功能 | TOPLINGDB_WITH_FABRICATED_COMPLEXITY
 
-**注意**: SidePlugin 暂不支持动态创建 ColumnFamily，混用 SidePlugin 和动态创建 ColumnFamily时，动态创建的 ColumnFamily 不能在 Web 中展示
-
 为了启用这些功能，需要为 make 命令显式添加 `EXTRA_CXXFLAGS="-D${MACRO_1} -D${MACRO_2} ..."`，例如编译带动态创建 ColumnFamily 的 rocksdbjava:
 ```
 make -j`nproc` EXTRA_CXXFLAGS='-DROCKSDB_DYNAMIC_CREATE_CF' rocksdbjava

README.md

@@ -1,21 +1,39 @@
 ## [中文版](README-zh_cn.md)
 ## ToplingDB: A Persistent Key-Value Store for External Storage
-ToplingDB is developed and maintained by [Topling Inc](https://topling.cn). It is built with [RocksDB](https://github.com/facebook/rocksdb). See [ToplingDB Branch Name Convention](https://github.com/topling/toplingdb/wiki/ToplingDB-Branch-Name-Convention).
+ToplingDB is developed and maintained by [Topling Inc](https://topling.cn). See [ToplingDB Branch Name Convention](https://github.com/topling/toplingdb/wiki/ToplingDB-Branch-Name-Convention).
 
-ToplingDB's submodule **[rockside](https://github.com/topling/rockside)** is the entry point of ToplingDB, see **[SidePlugin wiki](https://github.com/topling/rockside/wiki)**.
+## Quick Start
+ToplingDB requires C++17, gcc 8.3 or newer is recommended, clang also works.
+
+ToplingDB is forked form [RocksDB](https://github.com/facebook/rocksdb), much faster than RocksDB, try it by yourself:
+### Compile & run db_bench
+```bash
+sudo yum -y install git libaio-devel gcc-c++ gflags-devel zlib-devel bzip2-devel libcurl-devel liburing-devel snappy-devel jemalloc-devel
+#sudo apt-get update -y && sudo apt-get install -y libjemalloc-dev libaio-dev libgflags-dev zlib1g-dev libbz2-dev libcurl4-gnutls-dev liburing-dev libsnappy-dev libbz2-dev liblz4-dev libzstd-dev
+git clone https://github.com/topling/toplingdb
+cd toplingdb
+make -j`nproc` db_bench DEBUG_LEVEL=0
+sudo make install PREFIX=/some/path # default is /usr/local
+```
+
+After compile, you can run bundled [db_bench.sh](db_bench.sh)(need [port 2011](https://github.com/topling/rockside/blob/master/sample-conf/db_bench_enterprise.yaml#L4 "use port 2011 for embeded http server")), then use ToplingDB [in C++](https://github.com/topling/sideplugin-wiki-en/wiki/101 "maybe migrate from rocksdb"), or in [Java](https://github.com/topling/sideplugin-wiki-en/wiki/SidePlugin-Java-Binding "Bundled in this repo"), [Rust](https://github.com/topling/rust-toplingdb "A seperated repo").
+
+## Introduction
+ToplingDB's submodule **[rockside](https://github.com/topling/rockside)** is the entry point of ToplingDB, see **[SidePlugin wiki](https://github.com/topling/sideplugin-wiki-en/wiki)**.
 
 ToplingDB has much more key features than RocksDB:
-1. [SidePlugin](https://github.com/topling/rockside/wiki) enables users to write a json(or yaml) to define DB configs
-1. [Embedded Http Server](https://github.com/topling/rockside/wiki/WebView) enables users to view almost all DB info on web, this is a component of [SidePlugin](https://github.com/topling/rockside/wiki)
-1. [Embedded Http Server](https://github.com/topling/rockside/wiki/WebView) enables users to [online change](https://github.com/topling/rockside/wiki/Online-Change-Options) db/cf options and all db meta objects(such as MemTabFactory, TableFactory, WriteBufferManager ...) without restart the running process
+1. [SidePlugin](https://github.com/topling/sideplugin-wiki-en/wiki) enables users to write a json(or yaml) to define DB configs
+1. [Embedded Http Server](https://github.com/topling/sideplugin-wiki-en/wiki/WebView) enables users to view almost all DB info on web, this is a component of [SidePlugin](https://github.com/topling/sideplugin-wiki-en/wiki)
+1. [Embedded Http Server](https://github.com/topling/sideplugin-wiki-en/wiki/WebView) enables users to [online change](https://github.com/topling/sideplugin-wiki-en/wiki/Online-Change-Options) db/cf options and all db meta objects(such as MemTabFactory, TableFactory, WriteBufferManager ...) without restart the running process
 1. Many improvements and refactories on RocksDB, aimed for performance and extendibility
+1. memtable as wal log index, omit Flush MemTable to L0, reduce write amp, further improves for large MemTable
 1. Topling transaction lock management, 5x faster than rocksdb
 1. MultiGet with concurrent IO by fiber/coroutine + io_uring, much faster than RocksDB's async MultiGet
-1. Topling [de-virtualization](https://github.com/topling/rockside/wiki/Devirtualization-And-Key-Prefix-Cache-Principle), de-virtualize hotspot (virtual) functions, and key prefix caches, [bechmarks](https://github.com/topling/rockside/wiki/Devirtualization-And-Key-Prefix-Cache-Benchmark)
+1. Topling [de-virtualization](https://github.com/topling/sideplugin-wiki-en/wiki/Devirtualization-And-Key-Prefix-Cache-Principle), de-virtualize hotspot (virtual) functions, and key prefix caches, [bechmarks](https://github.com/topling/sideplugin-wiki-en/wiki/Devirtualization-And-Key-Prefix-Cache-Benchmark)
 1. Topling zero copy for point search(Get/MultiGet) and Iterator
 1. Topling memtable as log index, omit memtable flush to L0
 1. Builtin SidePlugin**s** for existing RocksDB components(Cache, Comparator, TableFactory, MemTableFactory...)
-1. Builtin Prometheus metrics support, this is based on [Embedded Http Server](https://github.com/topling/rockside/wiki/WebView)
+1. Builtin Prometheus metrics support, this is based on [Embedded Http Server](https://github.com/topling/sideplugin-wiki-en/wiki/WebView)
 1. Many bugfixes for RocksDB, a small part of such fixes was [Pull Requested](https://github.com/facebook/rocksdb/pulls?q=is%3Apr+author%3Arockeet) to [upstream RocksDB](https://github.com/facebook/rocksdb)
 
 ## ToplingDB cloud native DB services
@@ -48,38 +66,14 @@ toplingdb
 [ToplingDB](https://github.com/topling/toplingdb) | public | Top repository, forked from [RocksDB](https://github.com/facebook/rocksdb) with our fixes, refactories and enhancements
 [rockside](https://github.com/topling/rockside) | public | This is a submodule, contains:<ul><li>SidePlugin framework and Builtin SidePlugin**s**</li><li>Embedded Http Server and Prometheus metrics</li></ul>
 [cspp-wbwi<br>(**W**rite**B**atch**W**ith**I**ndex)](https://github.com/topling/cspp-wbwi) | public | With CSPP and carefully coding, **CSPP_WBWI** is 20x faster than rocksdb SkipList based WBWI
-[cspp-memtable](https://github.com/topling/cspp-memtable) | public | (**CSPP** is **C**rash **S**afe **P**arallel **P**atricia trie) MemTab, which outperforms SkipList on all aspects: 3x lower memory usage, 7x single thread performance, perfect multi-thread scaling)
-[topling-sst](https://github.com/topling/topling-sst) | public | 1. [SingleFastTable](https://github.com/topling/rockside/wiki/SingleFastTable)(designed for L0 and L1)<br/> 2. VecAutoSortTable(designed for MyTopling bulk_load).<br/> 3. Deprecated [ToplingFastTable](https://github.com/topling/rockside/wiki/ToplingFastTable), CSPPAutoSortTable
+[cspp-memtable](https://github.com/topling/cspp-memtable/blob/memtable_as_log_index/README_EN.md) | public | (**CSPP** is **C**rash **S**afe **P**arallel **P**atricia trie) MemTab, which outperforms SkipList on all aspects: 3x lower memory usage, 7x single thread performance, perfect multi-thread scaling)
+[topling-sst](https://github.com/topling/topling-sst) | public | 1. [SingleFastTable](https://github.com/topling/sideplugin-wiki-en/wiki/SingleFastTable)(designed for L0 and L1)<br/> 2. VecAutoSortTable(designed for MyTopling bulk_load).<br/> 3. Deprecated [ToplingFastTable](https://github.com/topling/sideplugin-wiki-en/wiki/ToplingFastTable), CSPPAutoSortTable
 [topling-dcompact](https://github.com/topling/topling-dcompact) | public | Distributed Compaction with general dcompact_worker application, offload compactions to elastic computing clusters, much more powerful than RocksDB's Remote Compaction
-[topling-rocks](https://github.com/topling/topling-rocks) | **private** | For build [Topling**Zip**Table](https://github.com/topling/rockside/wiki/ToplingZipTable), an SST implementation optimized for RAM and SSD space, aimed for L2+ level compaction, which uses topling dedicated searchable in-memory data compression algorithms
+[topling-rocks](https://github.com/topling/topling-rocks) | **private** | For build [Topling**Zip**Table](https://github.com/topling/sideplugin-wiki-en/wiki/ToplingZipTable), an SST implementation optimized for RAM and SSD space, aimed for L2+ level compaction, which uses topling dedicated searchable in-memory data compression algorithms
 [topling-zip_table_reader](https://github.com/topling/topling-zip_table_reader) | public | For read Topling**Zip**Table by community users, builder of Topling**Zip**Table is in [topling-rocks](https://github.com/topling/topling-rocks)
 
 To simplify the compiling, repo**s** are auto cloned in ToplingDB's Makefile, community users will auto clone public repo successfully but fail to auto clone **private** repo, thus ToplingDB is built without **private** components, this is so called **community** version.
 
-## Run db_bench
-ToplingDB requires C++17, gcc 8.3 or newer is recommended, clang also works.
-
-Even without ToplingZipTable, ToplingDB is much faster than upstream RocksDB:
-```bash
-sudo yum -y install git libaio-devel gcc-c++ gflags-devel zlib-devel bzip2-devel libcurl-devel liburing-devel snappy-devel jemalloc-devel
-#sudo apt-get update -y && sudo apt-get install -y libjemalloc-dev libaio-dev libgflags-dev zlib1g-dev libbz2-dev libcurl4-gnutls-dev liburing-dev libsnappy-dev libbz2-dev liblz4-dev libzstd-dev
-git clone https://github.com/topling/toplingdb
-cd toplingdb
-make -j`nproc` db_bench DEBUG_LEVEL=0
-cp sideplugin/rockside/src/topling/web/{style.css,index.html} ${/path/to/dbdir}
-cp sideplugin/rockside/sample-conf/db_bench_*.yaml .
-export LD_LIBRARY_PATH=`find sideplugin -name lib_shared`
-# change db_bench_community.yaml as your needs
-# 1. use default path(/dev/shm) if you have no fast disk(such as a cloud server)
-# 2. change max_background_compactions to your cpu core num
-# 3. if you have github repo topling-rocks permissions, you can use db_bench_enterprise.yaml
-# 4. use db_bench_community.yaml is faster than upstream RocksDB
-# 5. use db_bench_enterprise.yaml is much faster than db_bench_community.yaml
-# command option -json can accept json and yaml files, here use yaml file for more human readable
-./db_bench -json=db_bench_community.yaml -num=10000000 -disable_wal=true -value_size=20 -benchmarks=fillrandom,readrandom -batch_size=10
-# you can access http://127.0.0.1:2011 to see webview
-# you can see this db_bench is much faster than RocksDB
-```
 ## Configurable features
 For performance and simplicity, ToplingDB disabled some RocksDB features by default:
 
@@ -90,8 +84,6 @@ User level timestamp on key | TOPLINGDB_WITH_TIMESTAMP
 Wide Columns | TOPLINGDB_WITH_WIDE_COLUMNS
 fabricated features for read | TOPLINGDB_WITH_FABRICATED_COMPLEXITY
 
-**Note**: Dynamic creation of ColumnFamily is not supported by SidePlugin
-
 To enable these features, add `-D${MACRO_NAME}` to var `EXTRA_CXXFLAGS`, such as build ToplingDB for java with dynamic ColumnFamily:
 ```
 make -j`nproc` EXTRA_CXXFLAGS='-DROCKSDB_DYNAMIC_CREATE_CF' rocksdbjava

db/arena_wrapped_db_iter.cc

@@ -33,6 +33,38 @@ Status Iterator::RefreshKeepSnapshot(bool keep_iter_pos) {
   return Refresh(reinterpret_cast<Snapshot*>(KEEP_SNAPSHOT), keep_iter_pos);
 }
 
+Slice Iterator::NextWithKey() { return IterNextWithKeyImpl(this); }
+Slice Iterator::PrevWithKey() { return IterPrevWithKeyImpl(this); }
+
+Slice Iterator::SeekToFirstWithKey() {
+  SeekToFirst();
+  if (Valid())
+    return key();
+  else
+    return Slice(nullptr, 0);
+}
+Slice Iterator::SeekToLastWithKey() {
+  SeekToLast();
+  if (Valid())
+    return key();
+  else
+    return Slice(nullptr, 0);
+}
+Slice Iterator::SeekWithKey(const Slice& target) {
+  Seek(target);
+  if (Valid())
+    return key();
+  else
+    return Slice(nullptr, 0);
+}
+Slice Iterator::SeekForPrevWithKey(const Slice& target) {
+  SeekForPrev(target);
+  if (Valid())
+    return key();
+  else
+    return Slice(nullptr, 0);
+}
+
 ArenaWrappedDBIter::ArenaWrappedDBIter() {
   // do nothing
 }

db/arena_wrapped_db_iter.h

@@ -71,6 +71,9 @@ class ArenaWrappedDBIter final : public Iterator {
   }
   void Next() override { db_iter_->Next(); }
   void Prev() override { db_iter_->Prev(); }
+  Slice NextWithKey() override { return db_iter_->NextWithKey(); }
+  Slice PrevWithKey() override { return db_iter_->PrevWithKey(); }
+
   ROCKSDB_FLATTEN
   Slice key() const override { return db_iter_->key(); }
   ROCKSDB_FLATTEN

db/blob/blob_fetcher.h

@@ -40,7 +40,8 @@ class BlobFetcherCopyReadOptions : public BlobFetcher {
   const ReadOptions read_options_copy_;
 public:
   BlobFetcherCopyReadOptions(const Version* v, const ReadOptions& ro)
-   : BlobFetcher(v, read_options_copy_), read_options_copy_(ro) {}
+   : BlobFetcher(v, read_options_copy_),
+     read_options_copy_(ro, ReadOptions::BooleanDontCopyTrue()) {}
 };
 
 }  // namespace ROCKSDB_NAMESPACE

db/c.cc

@@ -1556,6 +1556,37 @@ void rocksdb_batched_multi_get_cf(rocksdb_t* db,
   delete[] statuses;
 }
 
+ROCKSDB_LIBRARY_API
+void rocksdb_batched_multi_get_cf_fast(rocksdb_t* db,
+                                  const rocksdb_readoptions_t* options,
+                                  rocksdb_column_family_handle_t* column_family,
+                                  size_t num_keys, const rocksdb_slice_t* keys_list,
+                                  rocksdb_pinnableslice_t** values, char** errs,
+                                  const bool sorted_input) {
+  PinnableSlice* value_slices = new PinnableSlice[num_keys];
+  Status* statuses = new Status[num_keys];
+
+  db->rep->MultiGet(options->rep, column_family->rep, num_keys, keys_list,
+                    value_slices, statuses, sorted_input);
+
+  for (size_t i = 0; i < num_keys; ++i) {
+    if (statuses[i].ok()) {
+      values[i] = new (rocksdb_pinnableslice_t);
+      values[i]->rep = std::move(value_slices[i]);
+      errs[i] = nullptr;
+    } else {
+      values[i] = nullptr;
+      if (!statuses[i].IsNotFound()) {
+        errs[i] = strdup(statuses[i].ToString().c_str());
+      } else {
+        errs[i] = nullptr;
+      }
+    }
+  }
+  delete[] value_slices;
+  delete[] statuses;
+}
+
 unsigned char rocksdb_key_may_exist(rocksdb_t* db,
                                     const rocksdb_readoptions_t* options,
                                     const char* key, size_t key_len,