Unanswered: High availability table design in V7 on OS/390
I am trying to architect a solution that maximizes availability with minimum impact on user applications that access a set of tables pretty much 24x7. These tables get refreshed overnight in batch mode and the process is spread over 3-5 hours. The idea of two seperate databases that are staggered in the refresh process was a no-go since they dont want the app developers to have to account for two dbs. Any solution that needs a small outage is OK. Based on browing through various sites and searching this forum for related topics, this is what I have gleaned so far. I want to run it by the forum as this is my first forray into the DB2 mainframe world having played primarily with Sybase/SQL Server on Solaris and Windows.
I am assuming a set of two similar tables will be staggered in the refresh and that once the new data is available the following options are viable:
1. Use Aliasing for the apps.
2. Use two sets of packages and switch
3. Use partitions and do a Alter table rename partition?
If I missed any other obvious options or if one of the above is a better implementation; feedback would be appreciated.
What do you mean with "get refreshed overnight in batch mode"? If you're talking about "loading" new data, I would consider using "online load" (SHRLEVEL CHANGE). Maybe you'll get some timeouts for the users transactions, but this is the simplest implementation (sometimes, simple is better ;-) ).
It is new data that gets loaded overnight. Will research your suggestion for the volume of data that we have.. Right now I am leaning towards package set switching using a control table which would dictate which set of tables to go for.