Database Reference
In-Depth Information
processing. Large-scope transaction processing is provided with optional support.
Partition and distribution are transparently operated, with client hash or a fixed
secret key space.
HyperTable was developed similar to BigTable to obtain a set of high-
performance, expandable, and distributed storage and processing systems for
structured and unstructured data [ 14 ]. HyperTable relies on distributed file systems,
e.g. HDFS, and distributed lock manager. Data representation, processing, and
partition mechanism are similar to that in BigTable. HyperTable has its own query
language, called HyperTable query language (HQL), and allows users to create,
modify, and query underlying tables.
Since the column-oriented storage databases mainly emulate BigTable, their
designs are all similar, except for the concurrency mechanism and several other
features. For example, Cassandra emphasizes weak concurrency of concurrent con-
trol of multiple editions, while HBase and HyperTable focus on strong consistency
through locks or log records.
4.3.1.3
Document Databases
Compared with key-value storage, document storage can support more complex data
forms. Since documents do not follow strict modes, there is no need to conduct mode
migration. In addition, key-value pairs can still be saved. We will examine three
important representatives of document storage systems, i.e., MongoDB, SimpleDB,
and CouchDB.
MongoDB
MongoDB is an open-source document-oriented database [ 15 ]. MongoDB stores
documents as Binary JSON (BSON) objects [ 16 ], which is similar to object. Every
document has an ID field as the main key word. Query in MongoDB is expressed
with syntax similar to JSON. A database driver sends the query as a BSON object to
MongoDB. The system allows query on all documents, including embedded objects
and arrays. Indexes may be created for queryable fields in documents to enable rapid
query.
The copy operation in MongoDB can be executed with log files in the main
nodes that support all the high-level operations conducted in the database. During
the copy operation, the machine queries all the writing operations since the last
synchronization of the machine and executing operations in log files in local
databases. MongoDB supports horizontal expansion with automatic sharing to
distribute data among thousands of nodes by automatically balancing load and keep
the system up and running in case of failure.
Search WWH ::




Custom Search