rocksdb/table/plain_table_reader.h
Lei Jin c83b085770 prefetch bloom filter data block for L0 files
Summary: as title

Test Plan:
db_bench
the initial result is very promising. I will post results of complete
runs

Reviewers: dhruba, haobo, sdong, igor

Reviewed By: sdong

Subscribers: leveldb

Differential Revision: https://reviews.facebook.net/D18867
2014-06-12 10:06:18 -07:00

261 lines
10 KiB
C++

// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
#pragma once
#ifndef ROCKSDB_LITE
#include <unordered_map>
#include <memory>
#include <vector>
#include <string>
#include <stdint.h>
#include "db/dbformat.h"
#include "rocksdb/env.h"
#include "rocksdb/iterator.h"
#include "rocksdb/slice_transform.h"
#include "rocksdb/table.h"
#include "rocksdb/table_properties.h"
#include "table/table_reader.h"
#include "table/plain_table_factory.h"
#include "util/arena.h"
#include "util/dynamic_bloom.h"
namespace rocksdb {
class Block;
class BlockHandle;
class Footer;
struct Options;
class RandomAccessFile;
struct ReadOptions;
class TableCache;
class TableReader;
class InternalKeyComparator;
using std::unique_ptr;
using std::unordered_map;
extern const uint32_t kPlainTableVariableLength;
// Based on following output file format shown in plain_table_factory.h
// When opening the output file, IndexedTableReader creates a hash table
// from key prefixes to offset of the output file. IndexedTable will decide
// whether it points to the data offset of the first key with the key prefix
// or the offset of it. If there are too many keys share this prefix, it will
// create a binary search-able index from the suffix to offset on disk.
//
// The implementation of IndexedTableReader requires output file is mmaped
class PlainTableReader: public TableReader {
public:
static Status Open(const Options& options, const EnvOptions& soptions,
const InternalKeyComparator& internal_comparator,
unique_ptr<RandomAccessFile>&& file, uint64_t file_size,
unique_ptr<TableReader>* table,
const int bloom_bits_per_key, double hash_table_ratio,
size_t index_sparseness, size_t huge_page_tlb_size);
Iterator* NewIterator(const ReadOptions&, Arena* arena = nullptr) override;
void Prepare(const Slice& target);
Status Get(const ReadOptions&, const Slice& key, void* arg,
bool (*result_handler)(void* arg, const ParsedInternalKey& k,
const Slice& v, bool),
void (*mark_key_may_exist)(void*) = nullptr);
uint64_t ApproximateOffsetOf(const Slice& key);
void SetupForCompaction();
std::shared_ptr<const TableProperties> GetTableProperties() const {
return table_properties_;
}
PlainTableReader(const Options& options, unique_ptr<RandomAccessFile>&& file,
const EnvOptions& storage_options,
const InternalKeyComparator& internal_comparator,
uint64_t file_size, const TableProperties* table_properties);
virtual ~PlainTableReader();
protected:
// Check bloom filter to see whether it might contain this prefix.
// The hash of the prefix is given, since it can be reused for index lookup
// too.
virtual bool MatchBloom(uint32_t hash) const;
// PopulateIndex() builds index of keys. It must be called before any query
// to the table.
//
// props: the table properties object that need to be stored. Ownership of
// the object will be passed.
//
// index_ contains buckets size of index_size_, each is a
// 32-bit integer. The lower 31 bits contain an offset value (explained below)
// and the first bit of the integer indicates type of the offset.
//
// +--------------+------------------------------------------------------+
// | Flag (1 bit) | Offset to binary search buffer or file (31 bits) +
// +--------------+------------------------------------------------------+
//
// Explanation for the "flag bit":
//
// 0 indicates that the bucket contains only one prefix (no conflict when
// hashing this prefix), whose first row starts from this offset of the
// file.
// 1 indicates that the bucket contains more than one prefixes, or there
// are too many rows for one prefix so we need a binary search for it. In
// this case, the offset indicates the offset of sub_index_ holding the
// binary search indexes of keys for those rows. Those binary search indexes
// are organized in this way:
//
// The first 4 bytes, indicate how many indexes (N) are stored after it. After
// it, there are N 32-bit integers, each points of an offset of the file,
// which
// points to starting of a row. Those offsets need to be guaranteed to be in
// ascending order so the keys they are pointing to are also in ascending
// order
// to make sure we can use them to do binary searches. Below is visual
// presentation of a bucket.
//
// <begin>
// number_of_records: varint32
// record 1 file offset: fixedint32
// record 2 file offset: fixedint32
// ....
// record N file offset: fixedint32
// <end>
Status PopulateIndex(TableProperties* props, int bloom_bits_per_key,
double hash_table_ratio, size_t index_sparseness,
size_t huge_page_tlb_size);
private:
struct IndexRecord;
class IndexRecordList;
// Plain table maintains an index and a sub index.
// index is implemented by a hash table.
// subindex is a big of memory array.
// For more details about the in-memory index, please refer to:
// https://github.com/facebook/rocksdb/wiki/PlainTable-Format
// #wiki-in-memory-index-format
uint32_t* index_;
int index_size_ = 0;
char* sub_index_;
const InternalKeyComparator internal_comparator_;
// represents plain table's current status.
Status status_;
Slice file_data_;
// data_start_offset_ and data_end_offset_ defines the range of the
// sst file that stores data.
const uint32_t data_start_offset_ = 0;
const uint32_t data_end_offset_;
const size_t user_key_len_;
const SliceTransform* prefix_extractor_;
static const size_t kNumInternalBytes = 8;
static const uint32_t kSubIndexMask = 0x80000000;
static const size_t kOffsetLen = sizeof(uint32_t);
static const uint64_t kMaxFileSize = 1u << 31;
static const size_t kRecordsPerGroup = 256;
// Bloom filter is used to rule out non-existent key
bool enable_bloom_;
DynamicBloom bloom_;
Arena arena_;
const Options& options_;
unique_ptr<RandomAccessFile> file_;
uint32_t file_size_;
std::shared_ptr<const TableProperties> table_properties_;
bool IsFixedLength() const {
return user_key_len_ != kPlainTableVariableLength;
}
size_t GetFixedInternalKeyLength() const {
return user_key_len_ + kNumInternalBytes;
}
friend class TableCache;
friend class PlainTableIterator;
// Internal helper function to generate an IndexRecordList object from all
// the rows, which contains index records as a list.
// If bloom_ is not null, all the keys' full-key hash will be added to the
// bloom filter.
Status PopulateIndexRecordList(IndexRecordList* record_list,
int* num_prefixes, int bloom_bits_per_key,
size_t index_sparseness);
// Internal helper function to allocate memory for indexes and bloom filters
void AllocateIndexAndBloom(int num_prefixes, int bloom_bits_per_key,
double hash_table_ratio,
size_t huge_page_tlb_size);
// Internal helper function to bucket index record list to hash buckets.
// bucket_header is a vector of size hash_table_size_, with each entry
// containing a linklist of IndexRecord hashed to the same bucket, in reverse
// order.
// of offsets for the hash, in reversed order.
// entries_per_bucket is sized of index_size_. The value is how many index
// records are there in bucket_headers for the same bucket.
size_t BucketizeIndexesAndFillBloom(
IndexRecordList* record_list, std::vector<IndexRecord*>* bucket_headers,
std::vector<uint32_t>* entries_per_bucket);
// Internal helper class to fill the indexes and bloom filters to internal
// data structures. bucket_headers and entries_per_bucket are bucketized
// indexes and counts generated by BucketizeIndexesAndFillBloom().
void FillIndexes(const size_t kSubIndexSize,
const std::vector<IndexRecord*>& bucket_headers,
const std::vector<uint32_t>& entries_per_bucket,
size_t huge_page_tlb_size);
// Read a plain table key from the position `start`. The read content
// will be written to `key` and the size of read bytes will be populated
// in `bytes_read`.
Status ReadKey(const char* row_ptr, ParsedInternalKey* key,
size_t* bytes_read) const;
// Read the key and value at `offset` to parameters `key` and `value`.
// On success, `offset` will be updated as the offset for the next key.
Status Next(uint32_t* offset, ParsedInternalKey* key, Slice* value) const;
// Get file offset for key target.
// return value prefix_matched is set to true if the offset is confirmed
// for a key with the same prefix as target.
Status GetOffset(const Slice& target, const Slice& prefix,
uint32_t prefix_hash, bool& prefix_matched,
uint32_t* offset) const;
Slice GetUserKey(const Slice& key) const {
return Slice(key.data(), key.size() - 8);
}
Slice GetPrefix(const Slice& target) const {
assert(target.size() >= 8); // target is internal key
return GetPrefixFromUserKey(GetUserKey(target));
}
inline Slice GetPrefix(const ParsedInternalKey& target) const;
Slice GetPrefixFromUserKey(const Slice& user_key) const {
if (!IsTotalOrderMode()) {
return prefix_extractor_->Transform(user_key);
} else {
// Use empty slice as prefix if prefix_extractor is not set. In that case,
// it falls back to pure binary search and total iterator seek is
// supported.
return Slice();
}
}
bool IsTotalOrderMode() const { return (prefix_extractor_ == nullptr); }
// No copying allowed
explicit PlainTableReader(const TableReader&) = delete;
void operator=(const TableReader&) = delete;
};
} // namespace rocksdb
#endif // ROCKSDB_LITE