First - a question - you say a client/server application stored on a DNS-323. Are you sure that you are, in fact, running a client/server application?
For this to happen, you would have had to hack the DNS-323 to get the server portion of the application running on it - you're going to need to install some sort of database server along with that application and given the limited memory and processing power of the DNS-323 - I'd say that would be quite a feat.
In a true client/server server scenario, the application is split into two portions - a client side, that runs on the client, and a server side, that runs on the server - the idea behind this is to reduce network traffic.
A relational database management system running on a computer and manipulating a database stored on a different computer will frequently generate significant volumes of network traffic - just as an example - a request to display all records with a particular description (show me all the 1TB drives in inventory), would require the workstation to search the entire database by requesting every record and examining the description field.
If this same system were designed as a client/server application - the client side application would send an SQL (structured query language) query to the server side application, which would then do the search, and return just the dozen or so records that match the search criteria.
A true client/server application will run using remote access with very little degradation, but the degradation you describe suggests that your application is not client/server but merely a multiple user RDBMS with a shared database.