By Margaret Hamburger-Oracle on Aug 14, 2013
Oracle TimesTen In-Memory Database release 126.96.36.199 is available for download. This release provides new capabilities that further improve performance and scalability for applications with extremely fast response time and very high throughput requirements, specifically for Online Transaction Processing (OLTP) workloads and Business Intelligence (BI) in-memory analytics workloads for Oracle Exalytics In-Memory Machine.
Feature highlights for Business Intelligence In-Memory Analytics workloads:
Performance and Scalability for In-Memory Analytics
The SQL Optimizer for in-memory analytics added a number of optimizations targeted for common star-join query patterns. These enhancements resulted in the optimizer choosing better query plans and speeding up overall query response time:
- Enables up to 3-10x query response time improvements for BI workloads and selected Oracle BI Applications;
- Enhanced in-memory query processing for star schemas
- Full Hash Index support. Prior to this release, hash indexes were supported only for primary key columns. Hash indexes may now be created on any set of columns to help speed up complex queries with equi-joins;
- Improved Index Advisor recommendations provide optimal indexes based on a specific query set.
Improvements to reduce data loading time
Data loading has been enhanced with the ability to support parallel, concurrent writes into a large compressed table. This feature significantly reduces the overall data loading time from the Data Warehouse to the Oracle TimesTen database in the Oracle Exalytics machine.
- Enables up to 3x reduction in data loading time for compressed tables;
- Improved data loading from the Oracle Database to TimesTen using the (ttLoadFromOracle) built-in function, with added support for parallel inserts to compressed tables;
- New utility (ttImportFromOracle) to analyze the data from an Oracle Database; the utility generates DDLs for tables and indexes with optimal Oracle TimesTen native data types and recommendation for columnar compression.
Feature highlights for Online Transaction Processing (OLTP) applications:
Optimistic Concurrent B+ Tree index for write scalability
A new B+Tree indexing capability featuring optimistic concurrency control provides increased scalability for write-intensive applications. This B+Tree index feature provides better throughput with higher numbers of concurrent writers, and is ideal for applications with very high volumes of insert and update transactions running in modern computing architectures with many processing cores.
- Greater concurrency support and higher transaction throughput for write intensive applications
- Enables up to 5x throughput improvement compared to existing B-tree indexes
Full Hash Index support for read scalability
The Hash index is a unique feature in Oracle TimesTen. A Hash index offers better performance for equality look up of Primary Key and is commonly used by OLTP application running on Oracle TimesTen databases. This release added the ability to create hash indexes on all columns in the table.
- Hash indexes can now be created for any single column and multiple-column indexes
- Enables up to 7x throughput improvement for read-intensive applications
Improved Scalability for NUMA systems
Modern computers are designed with high number of cores per server with large quantity of DRAM available. Such systems commonly exhibit Non Uniform Memory Access (NUMA) variations in memory latency, for example, when accessing memory outside of the local processor (socket). TimesTen as an in-memory database runs at memory (DRAM) speed and is more sensitive to the NUMA effect than a conventional disk-based relational database, where data access is optimized at the disk storage level. Improving TimesTen scalability on NUMA systems is an on-going effort in every release. A number of NUMA locality enhancements and code path optimizations were added in the 188.8.131.52 release, resulting in improved read and write scaling, especially on systems with very high number of processor cores such as the Oracle T5-8 servers, which have 128 cores and 1024 processing threads per server.
- 59.9 Million reads per second on Oracle T5-8 machine
- 1.03 Million update transactions per second on Oracle T5-8 machine
New Manageability Capabilities
New in this release is a utility for real-time monitoring and capturing snapshots of monitored metrics. This utility, ttStats, provides HTML reports of TimesTen database performance (similar to the Automatic Workload Repository (AWR) facility for the Oracle Database.) Also included in this new monitoring facility is a set of PL/SQL procedures for custom development. ttStats is a standalone TimesTen utility, invoked at the command line, and requires no additional software. ttStats can be used for real-time monitoring, snapshots captures, and HTML reports.
In addition to database level performance, the ability to time individual SQL statements is available in this release allowing the administrator to quickly identify long running operations.
Oracle TimesTen Quick Start Guide
A user-friendly, self-guided HTML Guide is available to get started with Oracle TimesTen. The Quick Start Guide uses the “How to” approach to provide easy to follow instructions for all components of the product, from database configurations, setups, to programming APIs, to configuring replication and cache grid, the instructions are provided step by step with screen shots and explanations for new and existing TimesTen users. The 184.108.40.206 Quick Start Guide also added many new Best Practices for Oracle TimesTen database and OS settings, Replication, Caching from the Oracle Database, and Monitoring.
Oracle TimesTen 220.127.116.11 release supports the following licensed products in the TimesTen product family: Oracle TimesTen In-Memory Database; Oracle TimesTen In-Memory Database for Exalytics; Oracle In-Memory Database Cache; and In-Memory Database Cache for Oracle Applications.