Suggestions wanted for new Project

[DeadManWalks]

Hi, we have an existing process on the iseries. I will call this the store forecast results build. This takes a long time due to the number of IOs that have to occur. This process has been changed to work on smaller batches of data at one time. This has reduced the time evolved greatly.

I suggested that we take this off line and move it to a pc. Basically the PC app will get all the data it needs up front from the 400 then churn away using just the resources of the PC. When done the "results" will then be sent to the 400.

I was thinking of doing this in VB using Access. I know the limitation of 2 gigs but I will be under this for each period that is run.

Comments are greatly appreciated.

[Billw]

Re: Suggestions wanted for new Project

What kind of high data reads are we talking about? all the months detail transactions or something like that. Could the process of detail transaction not update a montlhy summary for the transaction type (SKU number/store or whatever) accuming the quantity or accuming whatever. This way you are not dealing with all the individual transactions, just a summary of the transactions.

Used to do that with sales figures many years ago. Add to a summary file by product type for monthly/yearly figures. That way we didn't need to read 100,000s of records to get the information.

Pumping the data to a PC to do the work is not my idea of a controllable environment. Too many things can screwup.

Just some thoughts. Of course you know the process alot better than us.

Bill

[DeadManWalks]

Re: Suggestions wanted for new Project

At month end the sales of each store are combined with the "out" sales for the same item for the store.

next, 3 different forecasts are produced for this hisorical demand. A warehouse, A Model (groups of like stores), and store forecast.

The model and warehouse forecasts are quick (PC based). These forecasts are then used to build the results(the new demand points based on what the forecast says and what the location has).

The store forecast build (pc based) is quick. Its the results( new deamand points based on what forecast says, and other factors) that takes time.