Need help to develop Plan Migrating DB2 on z/OS to DB2 UDBv9.5 on AIX 5.3
Hello all,
I need help developing and documenting a usable data migration plan that describes the strategy, preparation, and specifications for migrating/converting from current DB2v9 on z/OS to DB2 UDB on AIX environment. I looking for document examples of a fully treated plan used by someone who had done this kind of project. My data is going from the mainframe to AIX, what are the challenges and gotchas, if any, I should be aware of? Any major difference between the DB2 object on the mainframe or AIX that may "trip" a AIX newbie? Are there some links to where I can find more resources for this ind of project?
Any and all response will br highly appreciated.
Okonita
I need help developing and documenting a usable data migration plan that describes the strategy, preparation, and specifications for migrating/converting from current DB2v9 on z/OS to DB2 UDB on AIX environment. I looking for document examples of a fully treated plan used by someone who had done this kind of project. My data is going from the mainframe to AIX, what are the challenges and gotchas, if any, I should be aware of? Any major difference between the DB2 object on the mainframe or AIX that may "trip" a AIX newbie? Are there some links to where I can find more resources for this ind of project?
Any and all response will br highly appreciated.
Okonita
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
I actually looked around and there are no references or examples. This type of migration is not done alot. I would suggest that you talk to IBM about getting somebody to help you.
I know we tested with DB2 on a distributed platform (I can't remember if it was Windows or Linux) and performance was less than desirable. The process normally takes 1 hour on z/OS and the "distributed" people originally stated that with the processing power they had the process should take less 30 minutes and in a worse case situation it make take equal to that.
After about 6 hours and the process was less than half way done the test was stopped.
I know we tested with DB2 on a distributed platform (I can't remember if it was Windows or Linux) and performance was less than desirable. The process normally takes 1 hour on z/OS and the "distributed" people originally stated that with the processing power they had the process should take less 30 minutes and in a worse case situation it make take equal to that.
After about 6 hours and the process was less than half way done the test was stopped.
Hi giltjr,
I was also surprised by the lack of examples.
You moved from a Z/OS system to LUW? If your tests suggest that LUW was taking about 10 times longer, it wasn't DB2, per se. There was a configuration issue, memory management, I/O paths, or some other peripheral element that was responsible for the degradation.
Kent
I was also surprised by the lack of examples.
You moved from a Z/OS system to LUW? If your tests suggest that LUW was taking about 10 times longer, it wasn't DB2, per se. There was a configuration issue, memory management, I/O paths, or some other peripheral element that was responsible for the degradation.
Kent
Not too many companies migrate DB2 from z/OS to distributed platforms. If they are migrating off of z/OS, they typically will go to a "cheaper" DBMS, like MySQL or PostgreSQL, or to DBMS that are considered better in the distributed world like Oracle.
The biggest issue was I/O.
The distributed team severely underestimated the I/O capability of our mainframe and severely overestimated the I/O capabilities of their SAN. They did not really understand the native multipath support that mainframe environment has has for decades and what the lack of such support would do on the distributed side.
They also over estimated the "advantage" of having more RAM and more faster CPU's. The mainframe at that time only has 10GB and they out 64GB on the distributed server. We had 5 CPU's rated at about 900Mhz vs. their 8 rated close to 3.0 Ghz.
The biggest issue was I/O.
The distributed team severely underestimated the I/O capability of our mainframe and severely overestimated the I/O capabilities of their SAN. They did not really understand the native multipath support that mainframe environment has has for decades and what the lack of such support would do on the distributed side.
They also over estimated the "advantage" of having more RAM and more faster CPU's. The mainframe at that time only has 10GB and they out 64GB on the distributed server. We had 5 CPU's rated at about 900Mhz vs. their 8 rated close to 3.0 Ghz.
There's always something to learn, huh?
Sorry that your lesson was so expensive.
Kent
ASKER