wbpesch@ihuxp.UUCP (Walt Pesch) (09/21/85)
Our group is currently designing an application that will use INGRES as our database manager. (More specifically, we are using AT&T INGRES, which when the new version is released will correspond to RTI's Version Three.) Each INGRES invocation creates a back-ground process. This sizes to about half a meg, with a quarter-meg as sharable text and the other quarter-meg as the data segment, unique for each process that has invoked INGRES. Because our application can have upwards of 100+ processes running on a dedicated 3B20 (most of which will want database access), our swapping rate for the background process would bring the machine to its knees. The normal response would be to have some background daemons that would feed off a message queue and do the database accesses for the rest of the processes. My question is what will be the effects of Paging Unix (specifically 5.2.2 - I hope that's the right name). Does anyone have any information as to the characterization of the page fault rates and locations for the background process? If INGRES is only normally bopping around say 20K instead of a quarter-meg, we can change our approach and have each process access INGRES. This would be allow us to eliminate a whole level of interfaces, and save both real-world and system time. I hope that someone out there can shed some light on the effects of Paging Unix on the data segment (especially you folks out at RTI). I am interested in anything pertaining to this, whether it be from 4.2 or whatever. Walt Pesch AT&T Network Systems ihnp4!ihuxp!wbpesch (312) 510-6237 Cornet 345-6237 -- Walt Pesch AT&T Network Systems ihnp4!ihuxp!wbpesch