Foreign Keys

Not Supported

Joins

Not Supported

Join operations are not supported based on the document.

Storage Model

N-ary Storage Model (Row/Record)

SSDB uses the storage model in LevelDB, where data is saved in sorted key-value pairs.

Query Execution

Tuple-at-a-Time Model

SSDB implements three separate iterator classes for key-value, hashmap, and sorted set.

Query Compilation

Not Supported

No query compilation is found in SSDB.

System Architecture

Shared-Nothing

SSDB is a single-node database management system. Initially, the server does not share anything with other nodes. However, as SSDB supports Redis API, developers can directly use Twemproxy to manage a number of SSDB servers. Besides, it is also workable to use the combination of Redis and SSDB.

Stored Procedures

Not Supported

Logging

Logical Logging

SSDB supports logical logging for its operations. It has multiple logging levels. The log file will change its name and create a new log file when exceeding the size limitation.

Storage Architecture

Disk-oriented

SSDB is a disk-oriented database. The key difference between Redis and SSDB is that Redis uses mainly in-memory storage while SSDB stores data in the disk. Therefore, SSDB can store 100 times data in Redis for each node.

Views

Not Supported

Concurrency Control

Not Supported

SSDB does not support concurrency control, because operations are executed individually, and there is no transaction. However, unlike Redis, SSDB supports multi-threads to avoid write operations blocking read operations (since 1.5.2). All write operations are executed in a single thread while read operations are executed in other threads.

Indexes

Skip List Log-Structured Merge Tree

SSDB uses LevelDB as its storage engine. SSDB explicitly calls LevelDB's interfaces to insert, update, and delete data. In LevelDB, a Log-Structured Merge Tree is used for indexes. Log-Structured Merge Tree provides a higher write throughput than the B+ tree. For each write operation, it first writes to a data structure called memtable, which will be flushed to disk later. In a memtable, LevelDB uses Skip List to find a position.

Query Interface

Custom API Command-line / Shell

Firstly, SSDB supports a set of shell commands to execute queries. Secondly, SSDB provides custom APIs for a wide range of programming languages, including C#, C++, Erlang, Go, Java, Lua, PHP, Python, Ruby, and Swift. Besides, developers can use the Redis client to connect the server with Redis Serialization Protocol. The command set of SSDB is similar to that in Redis.

Data Model

Key/Value

SSDB supports four main data types, which are key-value, hashmap, sorted set, and list. The key-value data type is used to store data with no strong relations. The hashmap is used to store data in different sets with counting and sorting based on the key value. Sorted set sorts all entries based on an integer of score. Unlike the key-value and hashmap data types, sorted set cannot be used to store large data. Besides, list data type stores data as a queue.

Compression

Dictionary Encoding

SSDB supports data compression using Snappy. Snappy is an open-source data compression library developed by Google. It is used in both Google's internal tools and several famous open-source databases, including Cassandra, MongoDB, and LevelDB. SSDB naturally supports Snappy as it uses LevelDB for low-level storage. In LevelDB, developers should explicitly set parameters to use Snappy in the compile time. Since version 1.6.2, SSDB has changed this feature by directly integrating the source code of Snappy to the project. By default, SSDB will use Snappy, but users can explicitly choose to use a non-compression version. SSDB with Snappy compression is compatible with non-compression data.

Checkpoints

Non-Blocking

SSDB provides the ssdb-dump tool for checkpointing and recovery. It also supports export and import commands to save and restore the database from the client side.

Storage Organization

Log-structured

SSDB uses the LevelDB storage, whose key organization is the log-structured merge tree.