Hello Baptiste: we were seeing lower tps from a remote system to the front-end LB hence trying to exclude client side issues by using the LB interface. Yes, when we use '-k', we do see a huge difference but its interesting that we pretty much always get 390 tps for a single core, and when we go to nbproc 2 then 780.
Appreciate the input Baptiste & Lukas. Thanks, Phil. ----- On 18 Jun, 2015, at 14:15, Baptiste bed...@gmail.com wrote: > Phil, > > First, use '-k' option on ab to keep connections alive on ab side. > > From a pure benchamrk point of view, using the loopback is useless! > Furthermore if all VMs are hosted on the same hypervisor. > You won't be able to get any accurate conclusion from your test, > because the injector VM is impacting the HAProxy VM, which migh be > mutually impacted the server VMs... > > Baptiste > > > On Thu, Jun 18, 2015 at 2:41 PM, Phil Daws <ux...@splatnix.net> wrote: >> Hello Lukas: >> >> Path is as follows: >> >> Internet -> HAProxy [Frontend:443 -> Backend:80] -> 6 x NGINX >> >> Yeah, unfortunately due to the application behind NGINX our benchmarking has >> to >> be without keep-alives :( >> >> Thanks, Phil >> >> ----- On 18 Jun, 2015, at 13:38, Lukas Tribus luky...@hotmail.com wrote: >> >>> Hi Phil, >>> >>> >>>> Hello all: >>>> >>>> we are rolling out a new system and are testing the SSL performance with >>>> some strange results. This is all being performed on a cloud hypervisor >>>> instance with the following: >>> >>> You are saying nginx listens on 443 (SSL) and 80, and you connect to those >>> ports directly from ab. Where in that picture is haproxy? >>> >>> >>> >>>> Have tried adding the option prefer-last-server but that did not make a >>>> great deal of difference. Any thoughts please as to what could be wrong ? >>> >>> Without keepalive it won't make any difference. Enable keepalive with ab >>> (-k). >>> >>> >>> >>> Lukas >> >> (null) (null)