Re[2]: CPU or NET VP Class ??

My problem is associated with this : When I use tlitcp for CPU VPs, then from my client machine, I can connect once to the database, and then if I try again, cannot connect. Hence I used shared memory for the NET VP and tcp for CPU VP. Which going by this discussion, will hit at the performance. What should I look at to identify the cause of only 1 connection and no more from the client m/c? Thanks Vandana vrchadha@afl.co.in ______________________________ Reply Separator _________________________________ Subject: Re: CPU or NET VP Class ?? Author: <kagel@bloomberg.net > at aflgate Date: 4/20/99 5:14 PM Lyzander Marantal wrote: > > In article <924603658.18749.1.nnrp-06.c2dec9bf@news.demon.co.uk>, > "Sean Kelsey" <seank@teepee.demon.co.uk> wrote: > > Good morning Informixers. > > > > So, okay, if we change the VP class to NET, will this not slow things down > > significantly for all user connections ? Can anyone point out what benefits > > (if any) changing the VP class will give us and which ones do you > > recommend... > Taken from Art Kagels Tech Notes Article: > "Do not configure shared memory NETTYPES to use NET VPs if possible. NET VPs > use polling for shared memory connections and that will spin your CPUs, tying > them up. Similarly avoid TCP NETTYPES using CPU VPs for similar reasons. It > increases the number of system calls per second and ties up OS services, > because the CPU VPs cannot sleep while waiting for new connections and > requests." Which BTW NET VPs will do when listening on TCP connections. > I've also changed my NETTYPE setting so that the network poll thread runs on a > NET VP and have seen a dramatic decrease in system calls. Try it. Good advice. Honest! > PS > I always treat Art's word as gospel and so far, everything has been smooth > sailing for me. Glad to have been of aid. ;-) Art S. Kagel