10 Ways to Leave your Upgrade, Faster

This blog is based on a RadiOAUG broadcast on April 9, 2013 at Collaborate in Denver.

10 Ways to Speed up your Upgrade.

1) Fast IO – Solid State Drive for the database files

2) Don’t do an in-place upgrade. This allows you to measure the upgrade on the actual machine you’ll use for the production upgrade.

3) Lots of disk space to stage files. Stage the 11i files so you don’t have to clone them before the upgrade. Stage the R1211 files and the post 1213 files, with all the customizations.

4) Try increasing parallel_max_servers to see if this allows greater parallelization.

5) Don’t use a threaded CPU architecture for the database server.

6) If you do use a threaded CPU server, use Automatic Degree of Parallelism (ADOP) to increase the parallelism. There are three modes: Manual, Limited and Auto. We tested Manual and Auto, and had the best results with Auto.

7) XDOLoader – entropy issue. Create a link to /dev/urandom called /dev/random. Note: 1065393.1.

8) Disable ARCHIVELOG Mode and Auditing

9) Use trace to diagnose performance issues:

oradebug setospid OSPID-number
oradebug unlimit
oradebug event 10046 trace name context forever, level 12
oradebug tracefile_name

Run tkprof on the tracefile and examine the tkprof output.

10) There are lots of functional scripts and datafixes that can speed your upgrade: facpupg.sql and glrsgup2.sql are two examples that allow you to run depreciation in advance and prepare posted journals before the upgrade.

Read more about it in my book, the little r12.1.3 tuning and troubleshooting guide for Oracle E-Business Suite