TimesTen In-Memory Database
for Extreme Performance

An Introduction to Oracle TimesTen Scaleout

By: Doug Hood | TimesTen Cloud Product Manager


Finally there is a scale out database where you don't need to trade data consistency for scalability. A database where you don't need to de-normalize to get extreme performance. An OLTP SQL database that is faster than NoSQL databases. This database is called TimesTen Scaleout.

Oracle TimesTen In-Memory Database has been a leading relational in-memory database for mission critical applications for over 20 years. Oracle TimesTen In-Memory Database has an exciting new scalability feature called TimesTen Scaleout. TimesTen Scaleout is a distributed database which has a shared nothing architecture suitable for low latency, high throughput OLTP, IoT and analytic workloads. TimesTen Scaleout supports high availability, ACID transactions, SQL & PL/SQL and does not require database de-normalization.

TimesTen Scaleout Use Cases


Scale Out

To enable linear scalability, TimesTen Scaleout uses a shared nothing architecture. This means that the database can scale horizontally over a set of machines. Adding more machines to a TimesTen Scaleout database enables more connections, greater throughput, larger capacity and higher availability. Adding more machines also enables SQL operations to perform in parallel. In general, the more machines the more effective the parallel SQL processing.

TimesTen Scaleout


High Availability

A TimesTen Scaleout database will usually span several machines. Multiple copies of the data can be configured so that there are no single points of failure. Having multiple copies of the data means that transient hardware and software failures do not affect the availability of the database. If all copies of the data become permanently un-available, then the database can be recovered from backup.

High Availability


ACID Transactions

Despite being an ‘In-Memory Database’, TimesTen databases are persistent across system or power failures. TimesTen persists data to a disk or flash filesystem.TimesTen Scaleout uses ACID (Atomic, Consistent, Isolated & Durable) transactions, just like Oracle databases.

TimesTen Scaleout achieves strong consistency via an optimized Two Phase Commit Protocol [2PC]. Unlike various 'eventual consistency' mechanisms, the 2PC protocol ensures that all changes are atomic. This strong consistency means that the transactional model is the same as if a single instance RDBMS is being used.  This means that developers do not need to worry about how transactions work, they just need to commit or rollback their units of work.

ACID Transactions


Single Database Image

Although the database can consist of many physical or virtual machines, it still appears to be a single logical database to developers and applications. Developers just see a single database with Schemas, Tables, Primary Keys, Foreign Keys, Indexes, Views, Materialized Views and Sequences.

Applications can modify any number of rows stored on any combination of machines in a single transaction.

Single DB Image


Table Distribution Clauses

As TimesTen is an In-Memory Database, it can execute extremely fast SQL Select/Insert/Update/Delete and Join operations within a single machine.

With other scale-out databases, as you spread your database across multiple machines you will tend to have performance problems. TimesTen Scaleout solves this by allowing you to co-locate commonly accessed data on the same machine. Depending on the SQL operation, data co-location can minimize or even eliminate network messaging and has a dramatic effect on performance and scalability.

Data co-location

Data co-location is configured via table distribution clauses. By default, rows in a table are evenly distributed across the machines in the TimesTen Scaleout database via consistent hash. For detail tables that are often joined, the foreign keys can be co-located with their parent table via the Distribute By Reference syntax. Reference tables which are 'read mostly' can be duplicated across all machines in a TimesTen Scaleout database via the Duplicate syntax.

Table distribution clauses



As TimesTen Scaleout uses a shared nothing architecture, it is possible to achieve linear scaleability for well tuned SQL workloads. At Oracle Open World 2017, a beta version of TimesTen Scaleout achieved 1.2 Billion SQL selects per second for 100% reads with the TPTBM workload.

1.2 Billion SQL Selects per second

This benchmark used 64 BM.HighIO1.36 compute nodes in the Oracle Cloud. By adding more Oracle Cloud compute nodes to the TimesTen Scaleout database the throughput continued to increase and this benchmark showed linear scaleability up to 64 machines.

144 million transactions per second


The same TPTBM workload also showed linear scalability with 80% reads and 20% writes achieving 144 million transactions per second using 64 database elements on 32 BM.HighIO1.36 compute nodes in the Oracle Cloud.

The Oracle Cloud Analytic Workload is a good example of a complex analytic workload [it uses the TCP-H schema and data but uses more complex queries] that scales as more machines are added to the TimesTen Scaleout database.

Scaleout to speed up analytics

This chart shows that as more machines are added to the TimesTen Scaleout database, the time to run all of the analytic queries is reduced. Specifically as the number of machines doubles, the latency halves. This dramatic reduction in latency is constrained by Amdahl's law.



TimesTen Scaleout supports SQL and PL/SQL via JDBC, ODBC, OCI, Pro*C, ttClasses and ODP.NET.

Supported SQL Drivers

TimesTen Scaleout also works with open source languages with SQL APIs that use an Oracle OCI driver or ODPI.

Open Source SQL Drivers


Developing Applications for TimesTen Scaleout

Existing TimesTen applications should work on TimesTen Scaleout with little or no change. Chongqing Mobile, [a subsidiary of China Mobile] recently went into production using TimesTen Scaleout by deploying an existing TimesTen application with only two changes. They changed the connect string to point to TimesTen Scaleout and they added distribution clauses to the tables to optimize the queries.

Also as TimesTen Scaleout uses a subset of Oracle Database compatible APIs, SQL, PL/SQL and data types, Oracle Database application developers can easily target their applications at TimesTen.

Develop Apps for TimesTen or TimesTen Scaleout

Details on how to develop applications for TimesTen Scaleout can be found here.


SQL Developer Support for TimesTen Scaleout

SQL Developer 18.2 supports full life cycle management for TimesTen Scaleout 18.1.

A SQL Developer window showing that the accounts table uses the distribute by reference table distribution clause.

SQL Deveoper Table DDL

Some SQL Developer windows showing a data model and the hash distribution of rows for the customers table.

SQL Developer Data Model and Row Distributions



TimesTen Scaleout is officially supported on these platforms and has been tested on the following Linux distributions:

Linux Distributions


TimesTen Scaleout is officially supported on Oracle VM and has been tested on the following VMs:

Virtual Machines



TimesTen Scaleout is software that runs on Linux x8664 hardware. The performance of TimesTen Scaleout is dependent on the performance of the hardware components - you get what you pay for. Dedicated bare metal machines will tend to give better performance than Virtual Machines.




  • Oracle TimesTen 18.1 Scaleout can be downloaded from here
  • Oracle SQL Developer 18.2 can be downloaded from here
  • A VirtualBox VM with a Hands On Lab for TimesTen 18.1 Scaleout can be downloaded from here
  • Sample code for TimesTen 18.1 Scaleout can be downloaded from here


Disclaimer: These are my personal thoughts and do not represent Oracle's official viewpoint in any way, shape, or form.

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.Captcha

Integrated Cloud Applications & Platform Services