Databases Reference
In-Depth Information
Oracle TimesTen
Oracle TimesTen is a relational database that is stored in physical memory and is typi‐
cally used where very high-performance transaction-processing workloads are present.
Access to the TimesTen database is available through SQL, JDBC, JMS, and ODBC.
TimesTen databases can be deployed as exclusive or shared and can be created as per‐
manent or temporary.
The database is refreshed by gathering data using TimesTen libraries deployed to ap‐
plications or by using a Cache Connect Option to an Oracle Database. Because data is
read and updated in memory, average update or read response times are typically meas‐
ured in the millionths of seconds. The Cache Connect Option supports both read and
write caching of Oracle Database data. Updates between TimesTen and Oracle can be
bidirectional. When paired with the Oracle Database, this is referenced as the Oracle
In-Memory Database Cache (IMDB Cache).
As is typical for embedded databases, TimesTen requires almost no ongoing adminis‐
tration. Replication is possible from one TimesTen database to another through an
option and is, by default, asynchronous.
Oracle introduced a variation named TimesTen for Exalytics with the Oracle Exalytics
In-Memory Machine in 2011. As might be expected for optimizing performance of the
BI Foundation Suite, key analytics and query functionality was added, including OLAP
grouping operators, analytic functions, time functions, and columnar compression.
Cloudera Distribution of Hadoop
Apache Hadoop is an open source framework for data-intensive applications where the
data is considered to be semi-structured or unstructured. Such data typically comes
from sensors, social media feeds, text, and web log data and contains descriptors, data
of value tied to those descriptors, and other miscellaneous data. Thus, the data of value
is relatively sparse. It was recognized by developers of search engines such as Google
and Yahoo! in the early 2000s that there was a need to map such data and reduce it down
to data of value to make sense of it. Hence, MapReduce was developed as a programming
paradigm and is embedded in Java, Python, and other programming languages. The
data itself is stored and analyzed in the Hadoop Distributed File System (HDFS) that is
deployed across a cluster of a multinode hardware configuration.
The most popular distribution of Hadoop, at the time this topic was published, is the
Cloudera Distribution of Hadoop (CDH). CDH is included with Oracle's Big Data Ap‐
pliance (BDA) and supported by Oracle. In addition to HDFS and MapReduce, CDH
includes other Hadoop components including Flume, Fuse-DFS, HBase, Hive, Mahout,
Oozie, Pig, Sqoop, and Zookeeper. CDH also provides the Cloudera Manager for man‐
aging the Hadoop cluster, and Oracle provides further platform management integra‐
tion via Enterprise Manager.
