Awesome
LesbianDB v2.1: high-performance, ACID-compilant NoSQL database
Getting started
You can get help on LesbianDB Server by running LesbianDB.Server --help.
Here's the list of valid arguments to pass to LesbianDB.Server at the time of writing.
--listen Required. The HTTP websocket prefix to listen to (e.g
https://lesbiandb-eu.scalegrid.com/c160d449395b5fbe70fcd18cef59264b/)
--engine Required. The storage engine to use (yuri/leveldb/saskia/purrfectodd)
--purrfectodd.flushinginterval (Default: 30000) Tells PurrfectODD to flush all writes to disk every N microseconds
--persist-dir The directory used to store the leveldb/saskia on-disk dictionary (required for
leveldb/purrfectodd, optional for saskia, have no effect for yuri)
--binlog The path of the binlog used for persistance/enhanced durability.
--soft-memory-limit (Default: 268435456) The soft limit to memory usage (in bytes)
--yurimalloc.buckets (Default: 65536) The number of YuriMalloc generation 1 buckets to create (only
useful for yuri storage engine, or saskia storage engine without --persist-dir
set).
--yurimalloc.gen2buckets (Default: 0) The number of YuriMalloc generation 2 buckets to create (only useful
for yuri storage engine, or saskia storage engine without --persist-dir set, zero
means YuriMalloc generation 2 disabled).
--yurimalloc.gen2promotiondelay (Default: 7200) The number of seconds to defer promotion of YuriMalloc data from
generation 1 to generation 2 (only useful for yuri storage engine, or saskia
storage engine without --persist-dir set, and YuriMalloc generation 2 is enabled).
--yuri.buckets (Default: 65536) The number of buckets to create (only used with Yuri storage
engine).
--accelerated-swap-compression (Default: disable) How should we use GPU-accelerated swap compression? disable: do
not use GPU-accelerated YuriMalloc swap compression, zram: use GPU-accelerated
memory compression as a replacement for swapping, zcache: hot data is stored in RAM
uncompressed, warm data is stored in RAM compressed, and cold data is swapped to
disk compressed. This feature requires NVIDIA CUDA compartiable GPUs.
--saskia.zram (Default: false) Tells the Saskia storage engine to use (in-CPU) memory compression
instead of YuriMalloc for swapping cold data (no effect if persist-dir is specified
or yuri/leveldb storage engine is used). This still has effect for PurrfectODD
since PurrfectODD uses saskia as it's cache.
--saskia.ephemeralbucketscount (Default: 65536) How many buckets should Saskia use in ephemeral mode (PurrfectODD
L2 cache or without --persist-dir)
--clear-binlog (Default: false) Clears the binlog on startup (only applictable for
PurrfectODD/Saskia storage engines). Make sure to backup your on-disk dictionary
first if you are using Saskia since Saskia cannot tolerate unexpected power
failures!
--no-read-cache (Default: false) Disables the read cache (useful for databases accessed solely via
the Optimistic Functions Framework, recognized by Saskia/PurrfectODD storage
engines)
--help Display this help screen.
--version Display version information.
In our examples, we will start an ephemeral database server with LesbianDB.Server --listen http://localhost:12345 --engine saskia --saskia.zram
Reads, writes, and conditional writes
Optimistic functions library
You may notice that the example above contains many optimization opportunilities and is a bit too complicated. To make life easier, we invented the optimistic functions library.
Here's an optimistic functions re-implementation of the first example. The optimistic functions library not only makes our life easier, but also performs many optimizations.
Database persistence
A database is not very useful if it's ephemeral. Luckily, we have binlog, on-disk dictionary, and hybrid persistence.
Binlog persistence
Binlog persistence uses an append-only file that logs all database writes. It can withstand unexpected power interruption and crashes very well, but the database can take forever to load. Binlog persistence is supported by all storage engines.
LesbianDB.Server --listen http://localhost:12345 --engine saskia --binlog mylittledatabase.binlog
On-disk dictionary persistence
On-disk dictionary persistence offers startup performance advantages at the cost of reduced durability against unexpected power interuption and crashes. It's not supported by the Yuri storage engine.
LesbianDB.Server --listen http://localhost:12345 --engine saskia --persist-dir mylittledatabase_persist
Hybrid persistence
Hybrid persistence combines on-disk dictionary persistence with binlog persistence for fast startup and durability.
LesbianDB.Server --listen http://localhost:12345 --engine leveldb --persist-dir mylittledatabase_persist --binlog mylittledatabase.binlog
Storage engine conversion
The binlog is useful if you want to switch storage engines. Hybrid and binlog persistence databases can be converted to a diffrent storage engine by simply loading the binlog with a new storage engine. Storage engine conversion for on-disk dictionary persistence and ephemeral databases are not supported yet.
IMPORTANT: YuriTables safe naming rule
YuriTables is a SQL-over-NoSQL overlay database built on top of LesbianDB v2.1. It is prone to certain forms of data corruption if not used properly.
Name collision
Creation of multiple YuriTables objects with the same name can lead to undefined behavior.
First underscore rule
2 YuriTables objects, or a YuriTable object and a "normal key" should not share the same prefix before the first underscore.
Unacceptable examples
jessielesbian_lesbiangirlfriends and jessielesbian_pets
jessielesbian and jessielesbian_lesbiangirlfriends
Acceptable examples
jessielesbian.lesbiangirlfriends and jessielesbian.pets
jessielesbian_lesbiangirlfriends and hillaryclinton_lesbiangirlfriends
jessie and jessielesbian