IBM Power Systems and IBM FlashSystem Flash Storage ... · PoC Performance Test with JD Edwards EnterpriseOne - Background Large medical manufacturing company JD Edwards EnterpriseOne
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Oracle Acceleration with IBM FlashSystemCustomer Challenge – Most Oracle applications are highly read-intensive. As such, additional processing power alone does little or nothing to improve performance. By placing all read data on low latency flash storage, reads will be performed much faster, boosting Oracle performance by up to 12x over conventional disk systems with no tuning or changes to code or system architecture.
IBM FlashSystem shared flash storage systems will►Decrease I/O wait time in mission critical Oracle
workloads ►Deliver the lowest latency available on any SAN
platform►Accelerate commonly I/O-bound workloads,
including transactional, batch and complex analytics
Power 795 server with 32 cores (POWER7) with 1 TB memory with AIX 6.1
SAN Volume Controller with 32 GB memory and 4 x 146GB SSDs
DS8300 storage system with approximately 8 TB of disk storage
*Note: Other than the logical partition running the database, core counts are approximate as the system is configured to run uncapped and dynamically move processing resource between the partitions as needed.
LPAR Cores* Memory GB Notes
Database 12 400 Oracle Database 11.1 (8 TB)
Applications 3 64 EnterpriseOne 9.0.2 with Tools 8.98.3.3 (900 users)
User Batch 2 32 EnterpriseOne 9.02 with Tools 8.98.3.3
Scheduled Batch 8 32 EnterpriseOne 9.02 with Tools 8.98.3.3
Web Servers (4) 1 24 Oracle Application Server 10.2 with Java™ 1.5 (32-bit)
Euro Apps 1 4 EnterpriseOne 9.0.2 with Tools 8.98.3.3 (50 users)
Euro User Batch 1 16 EnterpriseOne 9.0.2 with Tools 8.98.3.3
Euro Sched Batch 1 8 EnterpriseOne 9.0.2 with Tools 8.98.3.3
Nightly batch cycle consisting of 712 distinct jobs:►154 critical batch jobs including the MRP process
►134 non-critical JD Edwards EnterpriseOne batch jobs
►424 jobs pulling data from JD Edwards EnterpriseOne tables to populate a data warehouse
Also ran a single user test of the Sales Order Entry application using the Configurator function to determine interactive performance improvements, this was a subjective assessment
All Batch Critical batch Non-critical batch Data Warehouse jobs
Critical batch job processing (MRP process) run time reduced by 28%
Single user test of the Sales Order Entry application using the Configurator function determined interactive performance ran faster and was more responsive
The client uses Oracle E-Business Suite 12.1 applications to process their nightly order volumes. The cumulative order processing workflow consists of a number of job steps, those jobs are: ►High Volume Order Import►Pre Processor Wrapper Program
►Soft Check Wrapper Program
►Pre Allocation Wrapper Routing Program
►Regular Order Allocation Processor►Order Management Release to Warehouse Management System Wrapper
Program
►Post Release Wrapper Routing Program
The critical individual job steps for this client were the “High Volume Order Import” and “Regular Order Allocation Processor” jobs (in bold above)
The client was also interested in the “gather schema statistics” process performance for their production Oracle Database
IBM FlashSystem flash storage in combination with IBM Power Systems servers deliver significant performance benefits for real-world Oracle workloads
►Improving critical JD Edwards EnterpriseOne batch job (MRP process) run time by 28%
►Improving JD Edwards EnterpriseOne Sales Order Entry with Configurator interactive performance, making it run faster and be more responsive
►Helping the Oracle E-Business Suite “High Volume Order Import” job step run 4 times faster and “Regular Order Allocation Processor” job step run more than 2 times faster
►Boosting the Oracle Database Gather Schema Statistics histogram collection job to run 6 times faster