Re: ODBC Vs. Native connectivity Drivers

chaminda@jcdcrs4.jobcorps.org wrote: >I need some information on the difference between using odbc drivers and >native connectivity drivers. A lot of people have been arguing for some >time that odbc drivers are slower than native drivers. But I'd like to know actual >the reasons, if this is the case ?!! This area has been fairly well hashed through, you'll be happy to hear. Check out Ken North's articles on the subject (web search on "Ken North" and ODBC using the engine of your choice). He's a contributor to DBMS magazine, among others, and he's set up some downloadable client-server benchmarking programs, as well as statistics he's accumulated himself. To answer your question directly: ODBC 1.0 was s-l-o-w. Early ODBC drivers were poorly written. Two-tiered ODBC drivers (i.e. network-client-server) were originally pathetic, making one network RPC call per ODBC API call --- and the ODBC API usually requires the application to make a lot of calls to determine the full "personality" of the data source. But, raw query-response times are very good. In fact, some three-tiered ODBC systems significantly (like, by a factor of two) improve transfer performance, versus db vendors' own transports. Some things do change. Check out visigenic, simbatech, openlink and intersolv as 3rd-party sources for ODBC drivers and ODBC client-server systems. Go after some independent stuff like Ken North's studies. Get hold of some drivers (at least on free downloadable driver is on the simbatech web page) and run your own test. --------- Engineers believe that the equations approximate reality. Physicists believe that reality approximates the equations. Mathematicians never make the connection.