As many of you will know I am no stranger to the Upgrade and Unicode conversion (CUUC) process, having completed many of the them over the last ten years, my most recent being an MDMP conversion.
This time I have accepted a project which will stretch my technical expertise again, the data volumes involved are substantial - the client's Production R/3 database is 4TB in size, the BW system is over 2TB, the SRM system needs a two step upgrade process and the SCM system is split into three (Optimiser, Livecache, SCM application)
Thankfully I have a great team working with me, most of whom I have worked with previously on another CUUC project. So I have few worries about people being up to the challenge, and we all work in similar ways, which means we are all able to support each other easily and effectively.
The first challenge I had to work through was simply, how was the team going to work through such massive data volumes, when we had an effective window of 2.5days to accomplish our respective CUUC's. Some people will be screaming at their monitors with answers :-), but really the only sensible answer is to perform a parallel export and import. This is where the database is exported and written to a shared file system, once each package has completed it's file that file is made available to the import process - which then picks it up and starts importing it.
As you can imagine, the hardware requirements for such a project are extensive and so this is not a decision which can be taken lightly. So I decided to go out to some people for some 2nd opinions, I contacted technical account managers in SAP, HP (hardware vendor) and Microsoft (O/S & RDBMS vendor), everyone I spoke to advised the same thing - in order to shift that amount of data you must use parallel export and import.
During the bid process for this project, we found that the client was also planning to undergo a data centre migration, which suited me perfectly. The data centre migration, whilst it can complicate some things, also relieves pressure in one very key area - hardware contention.
Now things are falling into place, I have a method to execute the CUUCs and I have a project to help support my seemingly insatiable need for servers during this process. During my next few posts I will detail the various challenges and decisions of the project.