Big Data Storage - Big Data: Related Technologies, Challenges and Future Prospects - page 43

SimpleDB

SimpleDB is a distributed database and a web service of Amazon [ 17 ]. Data
in SimpleDB is organized into various domains in which data may be stored,
acquired, and queried. Domains include different properties and name/value pair
sets of projects. Date is copied to different machines at different data centers
in order to ensure data safety and improve performance. This system does not
support automatic partition and thus could not be expanded with the change of
data volume. SimpleDB allows users to use SQL to run query, e.g., selecting
sentences nonconforming to a single domain. It is worth noting that SimpleDB can
assure eventual consistency but does not feature MVCC (Muti-Version Concurrency
Control). Therefore, conflicts therein could not be detected from the client side.

CouchDB

Apache CouchDB is a document-oriented database written in Erlang [ 18 ]. Data
in CouchDB is organized into documents that consist of fields named by key
words/names and values, and are stored and accessed as JSON objects. Every
document is provided with a unique identifier. CouchDB allows access to database
documents through the RESTful HTTP API. If a document needs to be modified,
the client can download the entire document, modify it, and then send it back to
the database. After a document is rewritten once, the identifier will be modified and
updated. CouchDB utilizes the optimal copying to acquire scalability but without
a sharing mechanism. Since various CouchDBs may be executed along with other
transactions simultaneously, any kinds of Replication Topology can be built. The
consistency type of CouchDB relies on the copy mechanism. If the server-server
configuration is utilized, CouchDB system can ensure eventual consistency; with
the master-slave configure, strong consistency can be assured. MVCC in CouchDB
is synchronously executed with the historical Hash records.
Except for properties, sets, and indexes defined in sets, all documents are stored
without a document schema. Generally, they do not provide explicit lock and,
compared with traditional relational databases, feature weaker concurrency and
atomic properties. Documents may be distributed to nodes of all systems to achieve
scalability at different levels.

4.3.1.4 Platform for Nimble Universal Table Storage

Platform for Nimble Universal Table Storage (PNUTS) is a large-scale parallel
geographically-distributed system for Yahoo!'s web applications [ 19 ]. It relies on
a simple relational data model in which data is organized into a property record
table. In addition to the classic data types, blob (binary large object or basic large
object) is also an effective data type that allows any structures within records (not
always large-scale binary objects such as images or audio frequency).