D
dbeavon
Guest
I was able to confirm that sql92 servers use true multi-threading (they did so in 9.1D, long before the ABL remote servers): see: Progress KB - How to set up a secondary login broker for SQL-92 connections for Progress/OpenEdge database? "In Progress 9.1D and above the SQL engine is multithreaded" So given that the _sqlsrv2 process is multi-threaded, what would be wrong with configuring a process to allow 1000 user connections per server (-Ma 1000)? The OS itself probably has a reasonable limit to the number of threads per process, but is there an OE-specific reason to restrict the number of threads a lot more than the OS does? It seems to me that on the SQL92 side of things we might want to have a -Mpb value of just 2 or 3 (for redundancy) and then allow -Ma to get as high as the operating system allows. I don't have any of my own experiences tuning these parameters for SQL92. The docs don't make it clear why a person would prefer increasing -Ma over -Mpb (in a secondary SQL92 broker). (This question is about SQL92 and is NOT related to the ABL "remote servers". On the ABL side of things it is pretty clear why you would keep -Ma as low as possible since OE is trying to do its own CPU multi-tasking using a custom time-slicing algorithm and it will starve clients for CPU if the -Ma gets too high).
Continue reading...
Continue reading...