>Hi Doug...
>
>Sounds like a cool app.

Will be when it's done. *BG* I'm actually in the process of writing a little article about the COM stuff that I hope to get finished soon.

>I would take a different approach. Instead of being data down to a local cursor, I would create a VFP COM component that resides on the server that could be invoked via DCOM or remote automation. That COM server could use it's own local cursors to grab data from SQL, munge and do the analysis on the server. If need by, the resulting cursor could be thrown into an ADO recordset, and passed to the client.

I had thought about doing this. I've 448MB on the server but it's also my Exchange Server, SQL Server and everything else except firewall. I'm a little concerned about overworking it is why I haven't really pursued this approach. I'll have to watch the processor & memory load info more closely.

BTW, I do use a COM object on the server, driven from the VFP desktop (could be anything really) that calls a couple of SQL 7.0 SPs to do the heavy lifting. Very cool because when I need to log folks in from the Internet (to check on their account info etc) it's all ready to go. I return an ID from SQL through the COM object back to the desktop which serves as the User's unique logon ID for that session. I store that in an oSystem object property that gets squirted around as needed.

>
>Some time ago, I benchmarked indexing operations on server tables. The table was 150 mb's in size. The first scenario involved opening the table from a workstation. Indexing took about 10 minutes. The second scenario involved doing the indexing from a COM server on the server invoked with remote automation. Total time: 40 seconds....
>

No kidding... Mostly I'm going to get a copy of the bills, interest rates, current balances, etc and could easily run 1,000,000+ iterations. Typically there will be between 2-20 individual records and each with a couple of ways of payoffs. That will most likely end up with somewhere between 3(n-1) and 5(n-1) iterations. Having 10 credit cards and other unsecured debts @ 4 combinations results in 1,048,576 iterations if I've figured it correctly. Naturally I'll build in some logic branch avoidance as things proceed and I discover bad branches. I figure to get a couple three HOT boxes for this part. The records themselves are tiny compared to the processing requirement.

Having said that I can see a day where we provide this service on the Internet and then I'll process on the server. Well, actually, I'll build the engine to do it either way and drive it remotely.

I really won't need too much processing for indexes other than perhaps some sort of ranking mechanism. I guess after doing this for about 15 years that unless and until I can get a second server to offload the work to I'm a lot more comfortable with the local cursor approach but you have intrigued me here. I'm a little nervous about loading the one file server down. Any case studies or info you could direct me to?

Best,

DD