Performance of batch updates

, ,
1

We copy our browse programs from zOS to Win2003EE and see elapsed times of our jobs drop to 1/3 of the mainframe’s duration.

We copy our update programs from zOS to Win2003EE and see elapsed times of our jobs jump to 3 times the mainframe’s duration.

With MFSET=ON as our default, our first reaction was that Multi-fetch must be the cause of the slowdown, but we can see in Workbench that L6 and S4 calls in our test programs are not being multi-fetched.

  • Shouldn’t we expect the same performance improvement in our update batch jobs as in our reporting jobs?
    If so, what can we do to see those gains?
    If not, what can we do at least to match the mainframe’s performance?
    Is it our Adabas setup?
    Is it our Natparms?
    Has anyone seen and resolved this?

We don’t want to change any of the source code, other than to correct platform incompatibilities.

2

Hi Ralph;

The only thing that comes to mind is the Adabas Buffer Space. How do they compare on your original zOS system and the new WIN2003EE system. If the latter is a lot smaller than the zOS, and if you have relatively long running logical transactions, you might be incurring swapping overhead.

steve