SPARC T4-2 Produces World Record Oracle Essbase Aggregate Storage Benchmark Result
By Brian-Oracle on Aug 28, 2012
Significance of Results
Oracle's SPARC T4-2 server configured with a Sun Storage F5100 Flash Array and running Oracle Solaris 10 with Oracle Database 11g has achieved exceptional performance for the Oracle Essbase Aggregate Storage Option benchmark. The benchmark has upwards of 1 billion records, 15 dimensions and millions of members. Oracle Essbase is a multi-dimensional online analytical processing (OLAP) server and is well-suited to work well with SPARC T4 servers.
The SPARC T4-2 server (2 cpus) running Oracle Essbase 220.127.116.11.100 outperformed the previous published results on Oracle's SPARC Enterprise M5000 server (4 cpus) with Oracle Essbase 18.104.22.168 on Oracle Solaris 10 by 80%, 32% and 2x performance improvement on Data Loading, Default Aggregation and Usage Based Aggregation, respectively.
The SPARC T4-2 server with Sun Storage F5100 Flash Array and Oracle Essbase running on Oracle Solaris 10 achieves sub-second query response times for 20,000 users in a 15 dimension database.
The SPARC T4-2 server configured with Oracle Essbase was able to aggregate and store values in the database for a 15 dimension cube in 398 minutes with 16 threads and in 484 minutes with 8 threads.
The Sun Storage F5100 Flash Array provides more than a 20% improvement out-of-the-box compared to a mid-size fiber channel disk array for default aggregation and user-based aggregation.
The Sun Storage F5100 Flash Array with Oracle Essbase provides the best combination for large Oracle Essbase databases leveraging Oracle Solaris ZFS and taking advantage of high bandwidth for faster load and aggregation.
Oracle Fusion Middleware provides a family of complete, integrated, hot pluggable and best-of-breed products known for enabling enterprise customers to create and run agile and intelligent business applications. Oracle Essbase's performance demonstrates why so many customers rely on Oracle Fusion Middleware as their foundation for innovation.
(millions of items)
|SPARC T4-2, 2 x SPARC T4 2.85 GHz||1000||149||398*||55|
|Sun M5000, 4 x SPARC64 VII 2.53 GHz||1000||269||526||115|
|Sun M5000, 4 x SPARC64 VII 2.4 GHz||400||120||448||18|
* – 398 mins with CALCPARALLEL set to 16; 484 mins with CALCPARALLEL threads set to 8
128 GB memory
2 x 300 GB 10000 RPM SAS internal disks
SAS HBA with 2 SAS channels
Data Storage Scheme Striped - RAID 0
Oracle Solaris ZFS
Installer V 22.214.171.124.100
Oracle Essbase Client v 126.96.36.199.100
Oracle Essbase v 188.8.131.52.100
Oracle Essbase Administration services 64-bit
Oracle Database 11g Release 2 (184.108.40.206)
HP's Mercury Interactive QuickTest Professional 9.5.0
The benchmark test results include:
- Database Load: Time elapsed to build a database including outline and data load.
- Default Aggregation: Time elapsed to build aggregation.
- User Based Aggregation: Time elapsed of the aggregate views proposed as a result of tracked retrieval queries.
Summary of the data used for this benchmark:
- 40 flat files, each of size 1.2 GB, 49.4 GB in total
- 10 million rows per file, 1 billion rows total
- 28 columns of data per row
- Database outline has 15 dimensions (five of them are attribute dimensions)
- Customer dimension has 13.3 million members
- 3 rule files
Key Points and Best Practices
The Sun Storage F5100 Flash Array has been used to accelerate the application performance.
Setting data load threads (DLTHREADSPREPARE) to 64 and Load Buffer to 6 improved dataloading by about 9%.
Factors influencing aggregation materialization performance are "Aggregate Storage Cache" and "Number of Threads" (CALCPARALLEL) for parallel view materialization. The optimal values for this workload on the SPARC T4-2 server were:
Aggregate Storage Cache: 32 GB
Oracle Essbase Aggregate Storage Option Benchmark on
Oracle's SPARC T4-2 Server
- SPARC T4-2 Server
Oracle Database 11g Release 2 Enterprise Edition
Copyright 2012, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners. Results as of 28 August 2012.