User:Pkubat/BerkeleyDB alternatives

• libmdbx
  • Written in: C
  • Wrappers for:C
  • Packages available for: Not yet available in Fedora
  • Latest release: 0.8.2 / 2020-07-6
  • License: The OpenLDAP Public License
  • Search key length: minimum 0, maximum ≈¼ pagesize (1300 bytes for default 4K pagesize, 21780 bytes for 64K pagesize). Keys could be more than 2 times longer than LMDB.
  • Read-only access to DB: Supports concurrent readers.
  • Supported architectures:
  • Notes: libmdbx should be Compact and friendly for fully embedding. Only 25KLOC of C11, 64K x86 binary code, no internal threads neither processes, but implements a simplified variant of the Berkeley DB and dbm API.

• LMDB
  • Written in: C
  • Wrappers for: C++, Java, Python, Perl, Ruby, Go, Erlang, PHP, Lua, .NET, Common Lisp, node.js, Haskell, Tcl
  • Packages available for: Fedora, EPEL6+, RHEL8+
  • Latest release: 0.9.18 / 2016-02-05
  • License: OpenLDAP Public License
  • Search key length: Maximum length of 1/3 of a page (~1365B)
  • Read-only access to DB: The library is fully thread-aware and supports concurrent read/write access from multiple processes and threads. There is normally no pure read-only mode, since readers need write access to locks and lock file. (docs)
  • Supported architectures: i386/x86-64/PowerPC/ARM
  • Query keys larger than X: Using cursors: MDB_SET_RANGE - Position at first key greater than or equal to specified key.
  • Notes: Developed as a replacement for Berkeley DB in the OpenLDAP project - the API (while simplified) is very similar to BDB's. Optimized for read performance. B+Tree based.

• Kyoto Cabinet (replacing Tokyo Cabinet)
  • Written in: C++
  • Wrappers for: C, Java, Python, Ruby, Perl, and Lua
  • Packages available for: Fedora, EPEL5+
  • Latest release: 1.2.76 / 2012-05-24
  • License: GNU GPL
  • Search key length: No limit from database itself (uses c++ std::string or combination of c-style string and its length).
  • Read-only access to DB: See http://fallabs.com/kyotocabinet/spex.html, part Sharing One database by Multiple Processes. TL;DR: One DB cannot be accessed by multiple processes, DB file is protected by reader-writer lock (which means it will probably require write access even for read-only operation).
  • Supported architectures: Linux 2.6 and later (i386/x86-64/PowerPC/Alpha/SPARC)
  • Query keys larger than X: DB supports matching by regex, which should be possible to use for this kind of query. Other possibilities are match by prefix and match by similarity.
  • Notes: Records are organized in hash table or B+ tree.

• TDB
  • Written in: C
  • Wrappers for: Python
  • Packages available for: Fedora, RHEL6+
  • Latest release: 1.3.10 2016-07-28
  • License: GNU LGPL
  • Search key length: No limit apart from the data type used for holding the key and data lengths (int).
  • Read-only access to DB: Multiple processes can read a single database. Exclusion control is managed by a reader/writer lock. Internal locks in place to provide the ability for mutliple writer access (on different parts of the DB).
  • Supported architectures: i386/x86-64/ARM
  • Query keys larger than X:
  • Notes: Records are organized in hash table.

• GDBM
  • Written in: C
  • Wrappers for:
  • Packages available for: Fedora, RHEL6+
  • Latest release: 1.12 / 2016-05-16
  • License: GNU GPLv3
  • Search key length: No limit apart from the data type used for holding the key and data lengths (int).
  • Read-only access to DB: Multiple processes can read a single database. Exclusion control is managed by a reader/writer lock. Can be disabled by the user to provide his own way of exclusion control.
  • Supported architectures: i386/x86-64/ARM
  • Query keys larger than X:
  • Notes: Records are organized in hash table.

• LevelDB
  • Written in: C++
  • Wrappers for: C, Java, Go, Erlang, node.js
  • Packages available for: Fedora, EPEL6+
  • Latest release: 1.18 / 2014-09-16
  • License: 3-Clause BSD
  • Search key length: No limit apart from the data type used for holding the key length (size_t)
  • Read-only access to DB: A database may only be opened by one process at a time. The leveldb implementation acquires a lock from the operating system to prevent misuse. Within a single process, the same leveldb::DB object may be safely shared by multiple concurrent threads. (docs)
  • Supported architectures: i386/x86-64/PowerPC/ARM
  • Query keys larger than X: Native library's Iterator can Seek (Position at the first key in the source that at or past target) + reverse, node.js wrapper is able to use basic filters (greater than, lower than)
  • Notes: Does not support transactions. Records are organized in an LSM tree.

• RocksDB (Built on LevelDB)
  • Written in: C++
  • Wrappers for: C, Java (In development)
  • Packages available for: Not yet available in Fedora
  • Latest release: 4.9.0 / 2016-06-09
  • License: 3-Clause BSD
  • Search key length: No limit apart from the data type used for holding the key length (size_t)
  • Read-only access to DB: A database may only be opened by one process at a time. The rocksdb implementation acquires a lock from the operating system to prevent misuse. Within a single process, the same rocksdb::DB object may be safely shared by multiple concurrent threads. (docs)
  • Supported architectures: x86-64/PowerPC/ARM
  • Query keys larger than X: Iterator can Seek (Position at the first key in the source that at or past target) + reverse
  • Notes: Records are organized in an LSM tree.

• UnQlite
  • Written in: C
  • Wrappers for:
  • Packages available for: Not yet available in Fedora
  • Latest release: 1.1.6 / 2015-08-26
  • License: 2-Clause BSD
  • Search key length: No limit apart from the data type used for holding the key length (int). Data length limited by the data type unqlite_int64 (64b int).
  • Read-only access to DB: Supports concurrent readers. In order to be thread-safe, UnQLite must be compiled with the UNQLITE_ENABLE_THREADS directive defined.
  • Supported architectures:
  • Query keys larger than X: UNQLITE_CURSOR_MATCH_LE and UNQLITE_CURSOR_MATCH_GE when using cursors. This options have sense only if the underlying key/value storage subsystem support range search (i.e: B+Tree, R+Tree, etc.).
  • Notes: UnQLite works with run-time interchangeable storage engines (i.e. Hash, B+Tree, R+Tree, LSM, etc.).

Older (probably unmantained) projects

• TokyoCabinet (replacing QDBM)
  • Written in: C
  • Wrappers for: Perl, Ruby, Java, and Lua
  • Packages available for: Fedora, EPEL5+, RHEL6+
  • Latest release: 1.4.48 / 2012-08-18
  • License: GNU LGPL
  • Search key length: No limit apart from the data type used for holding the key and data lengths (int).
  • Read-only access to DB: Multiple processes can read a single database. Exclusion control is managed by a reader/writer lock.
  • Supported architectures: Linux 2.4 and later (x86-32/x86-64/PowerPC/Alpha/SPARC)
  • Query keys larger than X: It is possible to access each record with the cursor in ascending or descending order. According to this mechanism, forward matching search for strings and range search for integers are realized.
  • Notes: Records are organized in hash table, B+ tree, or fixed-length array.

• QDBM
  • Written in: C
  • Wrappers for: C++, Java, Perl, and Ruby
  • Packages available for: Fedora
  • Latest release: 1.8.78 / 2007-03-07
  • License: GNU LGPL
  • Search key length: No limit apart from the data type used for holding the key and data lengths (int).
  • Read-only access to DB: Multiple processes can read a single database. Exclusion control is managed by a reader/writer lock.
  • Supported architectures: Linux (2.2, 2.4, 2.6) (IA32, IA64, AMD64, PA-RISC, Alpha, PowerPC, M68000, ARM)
  • Query keys larger than X: It is possible to access each record with the cursor in ascending or descending order. According to this mechanism, forward matching search for strings and range search for integers are realized.
  • Notes: Records are organized in hash table or B+ tree.