Sorry but that is not the case. 2.8 pipelines requests if possible, and tries to maintain a pipeline to the back-end as well. 3.0 improves on that by attempting to maintain the connection to the back-end (if possible) between requests.
I suggest you check your testing method again. On Tue, 2020-03-24 at 09:04 +0000, Ankit Gupta wrote: > Hi, > > I had worked with the previous pound version 2.8. Through logs I can > see there is performance improvement opportunity: > > Lets say pound is installed in machine "B". > > (requester) (Pound installed) (response provider or server) > A ------------------ B ----------------------- P > > Machine "A" tries to send http request to machine "P", this goes > through "B". > > If machine A is trying to send lets say 1000 requests in a second it > is handled in pound by default 128 threads. > > These threads create a new connection with the machine "P" every > time. > > I tried to configure through pound.cfg with Alive, connTo,Client, > Timeout parameters but there is no such difference on the request > processing speed as new connections are being created. > > Can you provide a patch through which a single connection between "B" > and "P" can handle all the requests instead of creating a new http > connection every time? > > > > Regards, > Ankit Gupta > Mob No. :+91 7058580085 -- Robert Segall Apsis GmbH Postfach, Uetikon am See, CH-8707 Tel: +41-32-512 30 19 -- pound mailing list pound@apsis.ch https://admin.hostpoint.ch/mailman/listinfo/pound_apsis.ch
