You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: HISTORY.md
+24-7Lines changed: 24 additions & 7 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,8 +1,17 @@
1
1
# Speedb Change Log
2
2
3
3
## Unreleased
4
-
* move hashSpdb memtable from plugin to main code (#639)
5
-
* Support Spdb memtable in Java and C (#548)
4
+
5
+
### New Features
6
+
7
+
### Enhancements
8
+
9
+
### Bug Fixes
10
+
11
+
### Miscellaneous
12
+
13
+
## Grapes v2.6.0 (8/22/2023)
14
+
Based on RocksDB 8.1.1
6
15
7
16
### New Features
8
17
* Snapshot optimization - The most important information inside a snapshot is its Sequence number, which allows the compaction to know if the key-value should be deleted or not. The sequence number is being changed when modification happens in the db. This feature allows the db to take a snapshot without acquiring db mutex when the last snapshot has the same sequence number as a new one. In transactional db with mostly read operations, it should improve performance when used with multithreaded environment and as well other scenarios of taking large amount of snapshots with mostly read operations.
@@ -11,11 +20,9 @@
11
20
12
21
### Enhancements
13
22
* db_bench: add estimate-table-readers-mem benchmark which prints these stats.
14
-
15
23
* A new option on_thread_start_callback has been added. It allows to set thread affinity or perform other optimizations (e.g. NUMA pinning) to speedb background threads.
16
24
An example file on_thread_start_callback_example.cc has been provided to demonstrate how to use this feature.
17
-
18
-
25
+
* Support Spdb memtable in Java and C (#548)
19
26
20
27
### Bug Fixes
21
28
* unit tests: fix GlobalWriteControllerTest.GlobalAndWBMSetupDelay by waiting for the memtable memory release.
@@ -27,21 +34,28 @@ An example file on_thread_start_callback_example.cc has been provided to demonst
27
34
* unit tests: fix DBCompactionTest.DisableMultiManualCompaction by blocking all bg compaction threads which increased by default to 8 in #194.
28
35
* Proactive Flushes: fix accounting with non-WBM initiated flushes.
29
36
37
+
### Miscellaneous
38
+
* move hashSpdb memtable from plugin to main code (#639)
39
+
30
40
## Fig v2.5.0 (06/14/2023)
31
41
Based on RocksDB 8.1.1
32
42
33
43
### New Features
34
44
* Enable Speedb Features : Speedb users currently configure the database manually. New Speedb users are required to spend a lot of effort reading the documentation of the Speedb features.
35
45
The purpose of this feature is to help users enable and set Speedb options easily to a default configuration.
36
46
The SharedOptions class was added to improve the usability of multiple databases cases by arranging shared options.(#543)
37
-
38
-
### New Features
39
47
* Delay writes gradually based on memory usage of the WriteBufferManager (WBM).
40
48
Before this PR, setting allow_stall in the WBM's constructor meant that writes are completely stopped when the WBM's memory usage exceeds its quota. The goal here is to gradually decrease
41
49
the users write speed before that threshold is reached in order to gain stability.
42
50
To use this feature, pass allow_stall = true to the ctor of WBM and the db needs to be opened with options.use_dynamic_delay = true. The WBM will setup delay requests starting from (start_delay_percent * _buffer_size) / 100 (default value is 70) (start_delay_percent is another WBM ctor parameter).
43
51
Changes to the WBM's memory are tracked in WriteBufferManager::ReserveMem and FreeMem.
44
52
Once the WBM reached its capacity, if allow_stall == true, writes will be stopped using the old ShouldStall() and WBMStallWrites(). (#423)
53
+
* Prevent flush entry followed delete operations
54
+
currently during memtable flush , if key has a match key in the
55
+
delete range table and this record has no snapshot related to it,
56
+
we still write it with its value to SST file.
57
+
This feature keeps only the delete record and reduce SST size for later compaction.
58
+
(#411)
45
59
46
60
### Enhancements
47
61
* CI: add a workflow for building and publishing jar to maven central (#507)
@@ -66,18 +80,21 @@ Also switch to waiting a sec on the CV each time. This is required since a bg er
66
80
* db_bench: Create a WBM once for all db-s regardless of their use in different groups (#550)
67
81
* Tombstone unit test faiure (#560)
68
82
* build: Remove unused variables in unit tests (#581)
83
+
69
84
### Miscellaneous
70
85
* disable failing unit tests and paired bloom filter stress testing
71
86
* version: update Speedb patch version to 2.4.1 (#503)
72
87
73
88
## Speedb v2.4.1 ( 04/19/2023)
89
+
74
90
### Enhancements
75
91
* Add the ability to create any Filter Policy in java (including ribbon filter and the Speedb paired bloom filter) by @mrambacher in #387
76
92
77
93
### Bug Fixes
78
94
* Write Flow: Reduce debug log size. Note: the write flow is still experimental in this release (#461) by @ayulas in #472
79
95
80
96
## Ephedra v2.4.0 (04/05/2023)
97
+
81
98
### New Features
82
99
* New beezcli: Interactive CLI that offers data access and admin commands by @ofriedma in #427
83
100
* Global delayed write rate: manage the delayed write rate across multiple CFs/databases by @Yuval-Ariel in #392
0 commit comments