In-Memory and Exadata: A Potent Combination

May 31, 2019 | 2 minute read
Text Size 100%:

The Oracle Database supports a broad range of data formats, access methods, and programming languages under common administration, security policies, availability, and transactional consistency, for all kinds of workloads. For analytical workloads especially, in-memory technologies deliver dramatic speed and space efficiencies, and Exadata goes further with unique in-memory enhancements. In this post we discuss two Exadata-only in-memory features, including short videos where Shasank Chavan (VP, In-Memory Technologies) explains how they each work.

In-Memory Availability and Performance

Data stored in in-memory columnar (IMC) format are simply a columnar representation of data also stored in the traditional row format, and therefore enjoy the protections afforded regular database blocks, as the IMC store can always be reconstructed from the relevant blocks should. However, some applications cannot tolerate the performance impact of re-populating a node's IMC store, for example after a node crash. Exadata offers the ability to specify that their columnar data be duplicated to another database node, so that if node crashes, the application is transparently redirected to a surviving node.

In this 3 min video Shasank Chavan explains In-Memory fault tolerance features unique to Exadata. Duplicate mode provides for some or all of the data in a node's IMC store to be duplicated to another node. With full duplicate, a table can replicated in all the nodes, enabling even higher performance as all access to that table will be local to each node. 

Standby Database Extends In-Memory Capacity

As a pillar of database high availability, Active Data Guard complements its protection role with the ability to run workloads on a standby database node. Analytical reports are often offloaded to standby databases making them good candidates for using in-memory functionality. Exadata enables populating the in-memory store of the standby database with content different than those of the in-memory store of its corresponding primary database. 

In this 2 min video Shasank Chavan discusses Active Data Guard enhancements for In-Memory in Exadata that adds to the in-memory capacity of the primary by using the standby for different data. 

Exadata Delivers the Best Database In-Memory Experience

The Oracle database offers class-leading in-memory functionality that coexists with and enhances all other database functionality, without any compromises. Because it is designed and engineered by the database team, Exadata fully integrates in-memory technologies to deliver functionality that is simply unattainable on general-purpose systems. 

See also the In-Memory blog, the Maximum Availability Architecture blog, and our previous post on how high availability is built into Exadata. This paper discusses multimodel database management systems. For further technical material, see In-Memory, Data Guard, MAA, and Exadata sites.

We are always interested in your feedback. You are welcome to engage with us via Twitter @ExadataPM and by comments here.

Cristobal Pedregal-Martin

Cris Pedregal joined the Database Development organization at Oracle in 2005, where as a senior developer he contributed to the design and development of high availability functionality, the Recovery Appliance, compression technology, and the initial cloud deployments of the Oracle Database. He moved to the product management team, where he has managed and launched product features and evangelized Oracle technologies, and driven technical relationships with key external partners.

Prior to joining Oracle, Cris was on the Computer Science faculty at the University of New Mexico, Albuquerque. He holds a PhD in Computer Science at the University of Massachusetts, Amherst. He speaks one language with 2.5 accents.

 


Previous Post

How Oracle Builds Maximum Availability into Exadata

Next Post


Introducing Exadata Database Machine X8: The Foundation for Mission Critical On-Premises, Cloud, and Autonomous Databases

Gurmeet Goindi | 6 min read