Announcing RTD 188.8.131.52.11
By nicolasbonnet on Mar 02, 2012
It is our pleasure to let you know that Oracle just released a new version of the RTD platform addressing some important scalability requirements for high end deployments.
The driver for this new version, released as a patch on top of the RTD 184.108.40.206 platform was the need to address increasing volumes of “learnings” generated by ever-increasing volumes of “decisions”. This stems from the fact several of our high-end customers now have production environments with multiple cross-channel Inline Services supporting multiple “decisions” which make use of multiple predictive models using hundreds of data attributes to select from potentially thousands of “choices”. Addressing those high-end business requirements required increased RTD Learning Server capacity than was provided by RTD 220.127.116.11.
To address those needs, Oracle re-architected its RTD Learning Server engine to enable some level of parallelization. This new architecture relies on multi-threaded models updating and asynchronous learning records reading/deletions operations. This change provides a 150% improvement in learning record processing rates, which enables RTD to now process more than 58M incremental learning records per day with a deployment configuration consisting of 3 concurrently active inline services each with 900 choices, 200 data attributes, and 4 choice event predictive/learning models. This was achieved on a machine with 4 core / 6GB RAM allocated to the RTD learning server.
This new version of RTD is an important release for companies setting-up Big Data Analysis & Decision platforms in support of real-time and batch targeted customer experience enterprise deployments.
For complete details on this patch please refer to http://www.oracle.com/technetwork/middleware/real-time-decisions/psu11-1532856.html