2013-10-16 23:59:46 +02:00
|
|
|
// Copyright (c) 2013, Facebook, Inc. All rights reserved.
|
|
|
|
// This source code is licensed under the BSD-style license found in the
|
|
|
|
// LICENSE file in the root directory of this source tree. An additional grant
|
|
|
|
// of patent rights can be found in the PATENTS file in the same directory.
|
|
|
|
//
|
2013-12-11 04:03:13 +01:00
|
|
|
#include "db/memtable_list.h"
|
2012-11-29 01:42:36 +01:00
|
|
|
|
2014-10-28 19:54:33 +01:00
|
|
|
#ifndef __STDC_FORMAT_MACROS
|
|
|
|
#define __STDC_FORMAT_MACROS
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#include <inttypes.h>
|
2012-10-19 23:00:53 +02:00
|
|
|
#include <string>
|
2013-08-23 17:38:13 +02:00
|
|
|
#include "rocksdb/db.h"
|
2012-10-19 23:00:53 +02:00
|
|
|
#include "db/memtable.h"
|
2014-01-24 23:30:28 +01:00
|
|
|
#include "db/version_set.h"
|
2013-08-23 17:38:13 +02:00
|
|
|
#include "rocksdb/env.h"
|
|
|
|
#include "rocksdb/iterator.h"
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
2014-06-03 01:38:00 +02:00
|
|
|
#include "table/merger.h"
|
2012-10-19 23:00:53 +02:00
|
|
|
#include "util/coding.h"
|
2014-04-03 06:49:51 +02:00
|
|
|
#include "util/log_buffer.h"
|
2015-03-13 18:45:40 +01:00
|
|
|
#include "util/thread_status_util.h"
|
2012-10-19 23:00:53 +02:00
|
|
|
|
2013-10-04 06:49:15 +02:00
|
|
|
namespace rocksdb {
|
2012-10-19 23:00:53 +02:00
|
|
|
|
|
|
|
class InternalKeyComparator;
|
|
|
|
class Mutex;
|
|
|
|
class VersionSet;
|
|
|
|
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
MemTableListVersion::MemTableListVersion(MemTableListVersion* old) {
|
|
|
|
if (old != nullptr) {
|
|
|
|
memlist_ = old->memlist_;
|
|
|
|
size_ = old->size_;
|
|
|
|
for (auto& m : memlist_) {
|
|
|
|
m->Ref();
|
|
|
|
}
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
void MemTableListVersion::Ref() { ++refs_; }
|
|
|
|
|
2014-01-28 19:35:48 +01:00
|
|
|
void MemTableListVersion::Unref(autovector<MemTable*>* to_delete) {
|
2014-01-25 22:50:30 +01:00
|
|
|
assert(refs_ >= 1);
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
--refs_;
|
|
|
|
if (refs_ == 0) {
|
|
|
|
// if to_delete is equal to nullptr it means we're confident
|
|
|
|
// that refs_ will not be zero
|
|
|
|
assert(to_delete != nullptr);
|
|
|
|
for (const auto& m : memlist_) {
|
|
|
|
MemTable* x = m->Unref();
|
|
|
|
if (x != nullptr) {
|
|
|
|
to_delete->push_back(x);
|
|
|
|
}
|
2013-11-25 20:55:36 +01:00
|
|
|
}
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
delete this;
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
int MemTableListVersion::size() const { return size_; }
|
|
|
|
|
2012-10-19 23:00:53 +02:00
|
|
|
// Returns the total number of memtables in the list
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
int MemTableList::size() const {
|
|
|
|
assert(num_flush_not_started_ <= current_->size_);
|
|
|
|
return current_->size_;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Search all the memtables starting from the most recent one.
|
|
|
|
// Return the most recent value found, if any.
|
|
|
|
// Operands stores the list of merge operations to apply, so far.
|
|
|
|
bool MemTableListVersion::Get(const LookupKey& key, std::string* value,
|
2014-09-09 02:45:06 +02:00
|
|
|
Status* s, MergeContext* merge_context) {
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
for (auto& memtable : memlist_) {
|
2014-09-09 03:46:52 +02:00
|
|
|
if (memtable->Get(key, value, s, merge_context)) {
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
void MemTableListVersion::AddIterators(const ReadOptions& options,
|
2014-09-05 02:40:41 +02:00
|
|
|
std::vector<Iterator*>* iterator_list,
|
|
|
|
Arena* arena) {
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
for (auto& m : memlist_) {
|
2014-09-05 02:40:41 +02:00
|
|
|
iterator_list->push_back(m->NewIterator(options, arena));
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
In DB::NewIterator(), try to allocate the whole iterator tree in an arena
Summary:
In this patch, try to allocate the whole iterator tree starting from DBIter from an arena
1. ArenaWrappedDBIter is created when serves as the entry point of an iterator tree, with an arena in it.
2. Add an option to create iterator from arena for following iterators: DBIter, MergingIterator, MemtableIterator, all mem table's iterators, all table reader's iterators and two level iterator.
3. MergeIteratorBuilder is created to incrementally build the tree of internal iterators. It is passed to mem table list and version set and add iterators to it.
Limitations:
(1) Only DB::NewIterator() without tailing uses the arena. Other cases, including readonly DB and compactions are still from malloc
(2) Two level iterator itself is allocated in arena, but not iterators inside it.
Test Plan: make all check
Reviewers: ljin, haobo
Reviewed By: haobo
Subscribers: leveldb, dhruba, yhchiang, igor
Differential Revision: https://reviews.facebook.net/D18513
2014-06-03 01:38:00 +02:00
|
|
|
void MemTableListVersion::AddIterators(
|
|
|
|
const ReadOptions& options, MergeIteratorBuilder* merge_iter_builder) {
|
|
|
|
for (auto& m : memlist_) {
|
|
|
|
merge_iter_builder->AddIterator(
|
|
|
|
m->NewIterator(options, merge_iter_builder->GetArena()));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-04-23 02:17:33 +02:00
|
|
|
uint64_t MemTableListVersion::GetTotalNumEntries() const {
|
|
|
|
uint64_t total_num = 0;
|
|
|
|
for (auto& m : memlist_) {
|
|
|
|
total_num += m->GetNumEntries();
|
|
|
|
}
|
|
|
|
return total_num;
|
|
|
|
}
|
|
|
|
|
2014-01-27 02:40:43 +01:00
|
|
|
// caller is responsible for referencing m
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
void MemTableListVersion::Add(MemTable* m) {
|
|
|
|
assert(refs_ == 1); // only when refs_ == 1 is MemTableListVersion mutable
|
|
|
|
memlist_.push_front(m);
|
|
|
|
++size_;
|
|
|
|
}
|
|
|
|
|
2014-01-27 02:40:43 +01:00
|
|
|
// caller is responsible for unreferencing m
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
void MemTableListVersion::Remove(MemTable* m) {
|
|
|
|
assert(refs_ == 1); // only when refs_ == 1 is MemTableListVersion mutable
|
|
|
|
memlist_.remove(m);
|
|
|
|
--size_;
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Returns true if there is at least one memtable on which flush has
|
|
|
|
// not yet started.
|
2014-03-18 20:37:42 +01:00
|
|
|
bool MemTableList::IsFlushPending() const {
|
2013-09-05 02:24:35 +02:00
|
|
|
if ((flush_requested_ && num_flush_not_started_ >= 1) ||
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
(num_flush_not_started_ >= min_write_buffer_number_to_merge_)) {
|
2014-10-27 22:50:21 +01:00
|
|
|
assert(imm_flush_needed.load(std::memory_order_relaxed));
|
2012-10-19 23:00:53 +02:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2013-06-11 23:23:58 +02:00
|
|
|
// Returns the memtables that need to be flushed.
|
2014-01-02 20:26:57 +01:00
|
|
|
void MemTableList::PickMemtablesToFlush(autovector<MemTable*>* ret) {
|
2015-03-13 18:45:40 +01:00
|
|
|
AutoThreadOperationStageUpdater stage_updater(
|
|
|
|
ThreadStatus::STAGE_PICK_MEMTABLES_TO_FLUSH);
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
const auto& memlist = current_->memlist_;
|
|
|
|
for (auto it = memlist.rbegin(); it != memlist.rend(); ++it) {
|
2012-10-19 23:00:53 +02:00
|
|
|
MemTable* m = *it;
|
|
|
|
if (!m->flush_in_progress_) {
|
|
|
|
assert(!m->flush_completed_);
|
2012-11-29 01:42:36 +01:00
|
|
|
num_flush_not_started_--;
|
2012-10-19 23:00:53 +02:00
|
|
|
if (num_flush_not_started_ == 0) {
|
2014-10-27 22:50:21 +01:00
|
|
|
imm_flush_needed.store(false, std::memory_order_release);
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
2014-01-28 19:35:48 +01:00
|
|
|
m->flush_in_progress_ = true; // flushing will start very soon
|
2013-06-11 23:23:58 +02:00
|
|
|
ret->push_back(m);
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
|
|
|
}
|
2014-01-28 19:35:48 +01:00
|
|
|
flush_requested_ = false; // start-flush request is complete
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
|
|
|
|
2014-02-12 20:42:54 +01:00
|
|
|
void MemTableList::RollbackMemtableFlush(const autovector<MemTable*>& mems,
|
2014-11-07 20:50:34 +01:00
|
|
|
uint64_t file_number) {
|
2015-03-13 18:45:40 +01:00
|
|
|
AutoThreadOperationStageUpdater stage_updater(
|
|
|
|
ThreadStatus::STAGE_MEMTABLE_ROLLBACK);
|
2014-02-12 20:42:54 +01:00
|
|
|
assert(!mems.empty());
|
|
|
|
|
|
|
|
// If the flush was not successful, then just reset state.
|
|
|
|
// Maybe a suceeding attempt to flush will be successful.
|
|
|
|
for (MemTable* m : mems) {
|
|
|
|
assert(m->flush_in_progress_);
|
|
|
|
assert(m->file_number_ == 0);
|
|
|
|
|
|
|
|
m->flush_in_progress_ = false;
|
|
|
|
m->flush_completed_ = false;
|
|
|
|
m->edit_.Clear();
|
|
|
|
num_flush_not_started_++;
|
|
|
|
}
|
2014-10-27 22:50:21 +01:00
|
|
|
imm_flush_needed.store(true, std::memory_order_release);
|
2014-02-12 20:42:54 +01:00
|
|
|
}
|
|
|
|
|
2012-10-19 23:00:53 +02:00
|
|
|
// Record a successful flush in the manifest file
|
2013-06-11 23:23:58 +02:00
|
|
|
Status MemTableList::InstallMemtableFlushResults(
|
2014-10-02 01:19:16 +02:00
|
|
|
ColumnFamilyData* cfd, const MutableCFOptions& mutable_cf_options,
|
2015-02-05 06:39:45 +01:00
|
|
|
const autovector<MemTable*>& mems, VersionSet* vset, InstrumentedMutex* mu,
|
2014-11-07 20:50:34 +01:00
|
|
|
uint64_t file_number, autovector<MemTable*>* to_delete,
|
|
|
|
Directory* db_directory, LogBuffer* log_buffer) {
|
2015-03-13 18:45:40 +01:00
|
|
|
AutoThreadOperationStageUpdater stage_updater(
|
|
|
|
ThreadStatus::STAGE_MEMTABLE_INSTALL_FLUSH_RESULTS);
|
2012-10-19 23:00:53 +02:00
|
|
|
mu->AssertHeld();
|
|
|
|
|
|
|
|
// flush was sucessful
|
2013-10-28 06:55:46 +01:00
|
|
|
for (size_t i = 0; i < mems.size(); ++i) {
|
2013-06-11 23:23:58 +02:00
|
|
|
// All the edits are associated with the first memtable of this batch.
|
2013-10-28 07:08:54 +01:00
|
|
|
assert(i == 0 || mems[i]->GetEdits()->NumEntries() == 0);
|
2013-06-11 23:23:58 +02:00
|
|
|
|
2013-10-28 06:55:46 +01:00
|
|
|
mems[i]->flush_completed_ = true;
|
|
|
|
mems[i]->file_number_ = file_number;
|
2013-06-11 23:23:58 +02:00
|
|
|
}
|
2012-10-19 23:00:53 +02:00
|
|
|
|
|
|
|
// if some other thread is already commiting, then return
|
|
|
|
Status s;
|
|
|
|
if (commit_in_progress_) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Only a single thread can be executing this piece of code
|
|
|
|
commit_in_progress_ = true;
|
|
|
|
|
|
|
|
// scan all memtables from the earliest, and commit those
|
2013-06-11 23:23:58 +02:00
|
|
|
// (in that order) that have finished flushing. Memetables
|
|
|
|
// are always committed in the order that they were created.
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
while (!current_->memlist_.empty() && s.ok()) {
|
|
|
|
MemTable* m = current_->memlist_.back(); // get the last element
|
2012-10-19 23:00:53 +02:00
|
|
|
if (!m->flush_completed_) {
|
|
|
|
break;
|
|
|
|
}
|
2013-06-11 23:23:58 +02:00
|
|
|
|
2014-10-28 19:54:33 +01:00
|
|
|
LogToBuffer(log_buffer, "[%s] Level-0 commit table #%" PRIu64 " started",
|
|
|
|
cfd->GetName().c_str(), m->file_number_);
|
2012-10-19 23:00:53 +02:00
|
|
|
|
2012-11-29 01:42:36 +01:00
|
|
|
// this can release and reacquire the mutex.
|
2014-10-02 01:19:16 +02:00
|
|
|
s = vset->LogAndApply(cfd, mutable_cf_options, &m->edit_, mu, db_directory);
|
2012-10-19 23:00:53 +02:00
|
|
|
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
// we will be changing the version in the next code path,
|
|
|
|
// so we better create a new one, since versions are immutable
|
|
|
|
InstallNewVersion();
|
|
|
|
|
2013-06-11 23:23:58 +02:00
|
|
|
// All the later memtables that have the same filenum
|
|
|
|
// are part of the same batch. They can be committed now.
|
2013-10-25 00:58:00 +02:00
|
|
|
uint64_t mem_id = 1; // how many memtables has been flushed.
|
2013-06-11 23:23:58 +02:00
|
|
|
do {
|
|
|
|
if (s.ok()) { // commit new state
|
2014-10-28 19:54:33 +01:00
|
|
|
LogToBuffer(log_buffer, "[%s] Level-0 commit table #%" PRIu64
|
|
|
|
": memtable #%" PRIu64 " done",
|
|
|
|
cfd->GetName().c_str(), m->file_number_, mem_id);
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
current_->Remove(m);
|
2013-06-11 23:23:58 +02:00
|
|
|
assert(m->file_number_ > 0);
|
|
|
|
|
2013-11-27 23:56:20 +01:00
|
|
|
if (m->Unref() != nullptr) {
|
|
|
|
to_delete->push_back(m);
|
|
|
|
}
|
2013-06-11 23:23:58 +02:00
|
|
|
} else {
|
|
|
|
//commit failed. setup state so that we can flush again.
|
2014-10-28 19:54:33 +01:00
|
|
|
LogToBuffer(log_buffer, "Level-0 commit table #%" PRIu64
|
|
|
|
": memtable #%" PRIu64 " failed",
|
|
|
|
m->file_number_, mem_id);
|
2013-06-11 23:23:58 +02:00
|
|
|
m->flush_completed_ = false;
|
|
|
|
m->flush_in_progress_ = false;
|
|
|
|
m->edit_.Clear();
|
|
|
|
num_flush_not_started_++;
|
|
|
|
m->file_number_ = 0;
|
2014-10-27 22:50:21 +01:00
|
|
|
imm_flush_needed.store(true, std::memory_order_release);
|
2013-06-11 23:23:58 +02:00
|
|
|
}
|
2013-10-25 00:58:00 +02:00
|
|
|
++mem_id;
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
} while (!current_->memlist_.empty() && (m = current_->memlist_.back()) &&
|
2013-06-11 23:23:58 +02:00
|
|
|
m->file_number_ == file_number);
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
|
|
|
commit_in_progress_ = false;
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2012-11-29 01:42:36 +01:00
|
|
|
// New memtables are inserted at the front of the list.
|
2012-10-19 23:00:53 +02:00
|
|
|
void MemTableList::Add(MemTable* m) {
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
assert(current_->size_ >= num_flush_not_started_);
|
|
|
|
InstallNewVersion();
|
2014-01-27 02:40:43 +01:00
|
|
|
// this method is used to move mutable memtable into an immutable list.
|
|
|
|
// since mutable memtable is already refcounted by the DBImpl,
|
|
|
|
// and when moving to the imutable list we don't unref it,
|
|
|
|
// we don't have to ref the memtable here. we just take over the
|
|
|
|
// reference from the DBImpl.
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
current_->Add(m);
|
2013-08-23 08:10:02 +02:00
|
|
|
m->MarkImmutable();
|
2012-10-19 23:00:53 +02:00
|
|
|
num_flush_not_started_++;
|
|
|
|
if (num_flush_not_started_ == 1) {
|
2014-10-27 22:50:21 +01:00
|
|
|
imm_flush_needed.store(true, std::memory_order_release);
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Returns an estimate of the number of bytes of data in use.
|
|
|
|
size_t MemTableList::ApproximateMemoryUsage() {
|
2014-11-06 20:14:28 +01:00
|
|
|
size_t total_size = 0;
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
for (auto& memtable : current_->memlist_) {
|
2014-11-06 20:14:28 +01:00
|
|
|
total_size += memtable->ApproximateMemoryUsage();
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
2014-11-06 20:14:28 +01:00
|
|
|
return total_size;
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
|
|
|
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
void MemTableList::InstallNewVersion() {
|
|
|
|
if (current_->refs_ == 1) {
|
|
|
|
// we're the only one using the version, just keep using it
|
|
|
|
} else {
|
|
|
|
// somebody else holds the current version, we need to create new one
|
|
|
|
MemTableListVersion* version = current_;
|
|
|
|
current_ = new MemTableListVersion(current_);
|
2014-01-25 22:50:30 +01:00
|
|
|
current_->Ref();
|
MemTableListVersion
Summary:
MemTableListVersion is to MemTableList what Version is to VersionSet. I took almost the same ideas to develop MemTableListVersion. The reason is to have copying std::list done in background, while flushing, rather than in foreground (MultiGet() and NewIterator()) under a mutex! Also, whenever we copied MemTableList, we copied also some MemTableList metadata (flush_requested_, commit_in_progress_, etc.), which was wasteful.
This diff avoids std::list copy under a mutex in both MultiGet() and NewIterator(). I created a small database with some number of immutable memtables, and creating 100.000 iterators in a single-thread (!) decreased from {188739, 215703, 198028} to {154352, 164035, 159817}. A lot of the savings come from code under a mutex, so we should see much higher savings with multiple threads. Creating new iterator is very important to LogDevice team.
I also think this diff will make SuperVersion obsolete for performance reasons. I will try it in the next diff. SuperVersion gave us huge savings on Get() code path, but I think that most of the savings came from copying MemTableList under a mutex. If we had MemTableListVersion, we would never need to copy the entire object (like we still do in NewIterator() and MultiGet())
Test Plan: `make check` works. I will also do `make valgrind_check` before commit
Reviewers: dhruba, haobo, kailiu, sdong, emayanke, tnovak
Reviewed By: kailiu
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15255
2014-01-24 23:52:08 +01:00
|
|
|
version->Unref();
|
2012-10-19 23:00:53 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-10-04 06:49:15 +02:00
|
|
|
} // namespace rocksdb
|