Re: Performance problem on client/server

Byron, there was a similar issue recently. Your problem seems to be the network bottleneck, which is not an Informix problem itself. I have a setup with distributed machines, too, but I put all my batch- jobs onto the server, while all the user-interfaces are on the app- server. The reason is the amount of transferred date, and the number of users for an application. Suppose, there are two applications: one to handle the user input to your invoicing-system and the other to print out the invoices monthly. The former would be used by many persons entering comparably few data. The latter would be started once to process a large amount of collected data. This means many processes and low network consumption for the first -> app-server And one process and high data transfer for the latter -> same-maching (db-server) using shared-memory-connection At least, this is how i solved it. Sincerely, Roland Byron Rieper wrote: > > I am a newcomer to Informix (1 month) and have a performance issue in > a 4GL program I inherited. The site I am at has 3 DG machines running > Informix. Once is a standalone used for developement and the other > two are structured such that one is designated as a server with all > the data tables and the other contains all the programs. The program > I am working on prints 52 invoices in about 40 seconds on the > developement machine which has both test data and programs. When I run > the same 4GL program on the client with the same 52 invoices it takes > 50-60 minutes. How do I explain such a variance when the > client/server boxes are much bigger, faster and less in service than > the developement box at present.? My theory is that the way SQL > statements in the 4GL program are forcing the entire tables (about > 64000 rows) to be loaded down to client when executed. I have come to > this conclusion because I have seen similar types of programming abuse > in VB when failing to do a passthrough or using a stored procedure > against the server. Is this what is going on here? BTW I have done a > rlogin to the server and manually done the same SQL statements > embedded in the program and had the server return the population in > seconds; the selection criteria of all statements are entirely against > indexed columns. If the entire table is, in fact being uploaded to > the client every time then is the work around to use an SPL?