History[04]

In 2010, the origination of Couchbase, Membase, was developed by a company called NorthScale which was founded by developers from Memcached project.

In February, 2011, the Membase project founders and Membase, Inc. merged with CouchOne(a company with many engineers behind CouchDB) as a new company called Couchbase, Inc.

In January 2012, Couchbase Server 1.8 was released and 1.8 was the first version after rename from Membase.

In December 2012, Couchbase Server 2.0 was released. New features like JSON new document store, indexing and querying, incremental MapReduce and replication across data centers are added.

The latest version of Couchbase Server is 6.0.0.

Checkpoints[05]

The checkpoint inside the Couchbase Server is a record of last sequenceID. The Sequence ID is an internal unique number for every change happened in Couchbase Server with chronologically increasing order. The replicator of the Couchbase Server sends the checkpoint of last change to the target at the end of every replication cycle. The target here could be any remote database where the changes are replicated.

Compression[03]

Couchbase Server (Enterprise Edition) applies data compression to documents with a 3rd party open-source library Snappy. It provides three modes of compression.

• Off Mode. The Couchbase Server decompresses the document if it is compressed, and stores the uncompressed document in memory but recompresses it when storing on disk. Also, Couchbase server sends the document in uncompressed form.

• Passive Mode. When Couchbase Server receives the compressed document, it stores the compressed document in memory and disk. Also, the documents are sent in compressed form.

• Active Mode. Couchbase Server stores and sends compressed document even if the document is uncompressed.

Concurrency Control[06]

Couchbase Server offers both optimistic and pessimistic locking to guarantee concurrency. The optimistic locking is implemented on the Compare-and-Swap (CAS) Value, which is a unique and atomically incrementing identifier inside the metadata of document. Couchbase Server verifies the CAS Value before a document is modified. Couchbase Server also supports pessimistic locking but it is less commonly used.

Data Model[07]

The documents stored in Couchbase Server are in JSON or binary format.

Foreign Keys[08]

As a key-value database, there is no foreign key in the Couchbase Server.

Hardware Acceleration[09]

Not Supported

Indexes[10][11]

A few of indexes are available in Couchbase Server:

• Primary Index is based on the unique key of every item in a specified bucket. The bucket contains a full set of keys in a given key space.

• Global Secondary Index (GSI) is the most frequently used index in Couchbase Server, for queries performed with the N1Ql query-language. The indexer for GSI creates a B+ tree for fast scans on the index key. Also, there is a Memory-Optimized GSI index which uses the skip-list structure as opposed to B-tree indexes.

• Full Text is provided by the Search Service and it contains targets derived from the contents of documents within one or more specified buckets.

• Views are supported with fields and information extracted from documents.

Isolation Levels[12]

Couchbase Server provides read committed isolation by default at the single document level. However, by using the provided optimistic locking and pessimistic locking, user can obtain serialized isolation of an individual document.

Joins[13]

Couchbase Server supports lookup join, index join and ANSI join in its N1QL language.

Query Interface[14]

Couchbase Server presents a new language called N1QL which is a SQL extension for JSON as the query interface.

Storage Architecture[15]

Couchbase Server maintains recent accessed data in the memory to achieve high performance and scalability. The data which is not recently used will be swapped on to disks.

Storage Model[16][17]

Couchbase Server stores items in the Buckets. There are two kinds of Buckets in the Couchbase. One is Couchbase Buckets which could reside both in memory and disks. Another is Ephemeral bucket which only stays in the memory.

Storage Organization[17]

Couchbase Server uses Couchstore to store core data operations. Each Bucket (the smallest unit of Couchbase's storage model) is stored as a separate Couchstore file in the file system. Couchstore uses a B+tree structure to quickly access items through their keys. Append-only write model is used for efficient and safe writes.

Stored Procedures[18]

A similar feature called "Prepared queries" is supported in Couchbase Server.

System Architecture[19]

Each machine running Couchbase server is a single node, and multiple nodes could join together as a cluster. Every node of the cluster has a Cluster Manager module for communication. The data inside the cluster is evenly distributed and replicated over the nodes.

Views[20][21]

Three views are supported by Couchbase Server: MapReduce views, Spatial views and Global Secondary Indexes (GSI) views. They could be accessed by customized APIs or N1QL queries.