Loading ...
Sorry, an error occurred while loading the content.

33863Re: [LoadRunner] RE: Re: TruClient Load Test

Expand Messages
  • John Crunk
    Sep 18, 2013
    • 0 Attachment
      Would you have any on making customizations to Analysis Templates? HP apparently doesn't share these!

      Sent from my iPhone

      On Sep 18, 2013, at 10:46 AM, "James Pulley" <loadrunner-lists@...> wrote:

       

      Here ya go.   Courtesy of Mr. Bishop at Trust IV in the UK.   Original Sourcing for the VMWARE integration information: Petar Puskarich and his conversations with HP

       

      http://blog.trustiv.co.uk/2013/06/using-virtual-machines-load-generators

       

      From: LoadRunner@yahoogroups.com [mailto:LoadRunner@yahoogroups.com] On Behalf Of Matthew Stephenson
      Sent: Wednesday, September 18, 2013 9:42 AM
      To: LoadRunner@yahoogroups.com
      Subject: Re: [LoadRunner] RE: Re: TruClient Load Test

       




      Good Morning James!

       

      Is there any documentation from HP on the integration with VMWARE?   Google is not my friend today.

       

      Thanks,

       

      Matt

       


      From: "loadrunner-lists@..." <loadrunner-lists@...>
      To: LoadRunner@yahoogroups.com
      Sent: Wednesday, September 11, 2013 2:59 PM
      Subject: [LoadRunner] RE: Re: TruClient Load Test

       

       

      Actually there is a nice integration with VMWARE at the load generator software level that now allows access to the core VMWARE host clock instead of the floating clock in the VM for the virtual users, which was the root of much of my displeasure with running virtual users inside of virtual machines.  Makes the whole funky cock float issues disappear from the timing records.   It doesn't address the whole uncontrolled initial and in test conditions issues from cross VM shared resources (network, disk, ....), but it does at least firm of the integrity of the timing records which is a very good thing.

       

      In the case of EC2, I would never use those hosts for stats on performance, I would look to my control group running on physical hardware inside of my domain for the reference data.   The remote EC2 hosted items would make good noise generators for the test and pretty cheap too, as long as you remember to shut them down.   It would just be nice if HP provided a way to access our Amazon|Azure|Rackspace|.... account within LoadRunner/Performance center and then have access to an AMI specifically tuned for load generation with a matching patch level to my controller (HP Hint Hint Hint Hint)

       

      Which is a worst practice, running truclient on the controller or running Truclient on VMWARE?   Definitely running on the controller in my book.

       

      :)

       

      Or, maybe I have just been hanging around with Tomlinson too long.   



      --- In LoadRunner@yahoogroups.com, <loadrunner@yahoogroups.com> wrote:

      " Have you considered Amazon web services for EC2 load
      generators? Granted these would be outside of your physical domain.
      If you have a physical machine that you can use as a control element then perhaps you can consider virtual machines in your internal cloud (Vmware, Microsoft, xen, etc.) to provision some load generators for your use."

      Did Mark Tomlinson Hack Pulley's Account?

      http://www.sqaforums.com/showthreaded.php?Number=658826

      --- In LoadRunner@yahoogroups.com, "James Pulley" <loadrunner-lists@...> wrote:
      >
      > Best Practices dictates
      >
      > . Don't run any users on the controller, independent of protocol or
      > tool. The controller is busy enough and you don't want to distort the
      > performance of your virtual users by constantly having to service the video
      > interface on a load generator. Video takes higher priority to service
      > than does just "App" in the OS
      >
      > . Always have at least three load generators, hardware matched. Two
      > for primary load and one for a control set
      >
      >
      >
      > Yes, you will get far fewer TruClient than you will get HTML Mode users
      > which would be fewer than URL mode and far fewer than Winsock on the same
      > load generator. As you move up the stack your virtual users get heavier
      > and heavier resource wise. That is the trade off in moving your labor
      > quotient out of your development phase you are going to pay for the utility
      > increase in labor by shifting your costs into your load generators. It
      > may not be a one:one for swap, but the challenge will be there.
      >
      >
      >
      > Benchmark your load generators. Because of the way the code is marked as
      > swappable you can do things which will negatively impact your performance on
      > the load generators. Never go above 80% on CPU or RAM. Watch your swap
      > levels as this would indicate that you are oversubscribing the OS. Some
      > people will get four per box and others 40, depending on hardware and
      > environment setup. Have you considered Amazon web services for EC2 load
      > generators? Granted these would be outside of your physical domain.
      > If you have a physical machine that you can use as a control element then
      > perhaps you can consider virtual machines in your internal cloud (Vmware,
      > Microsoft, xen, etc.) to provision some load generators for your use.
      >
      >
      >
      >
      >
      >
      >
      > From: LoadRunner@yahoogroups.com [mailto:LoadRunner@yahoogroups.com] On
      > Behalf Of Stimely, Noelle
      > Sent: Wednesday, September 11, 2013 10:17 AM
      > To: LoadRunner@yahoogroups.com
      > Subject: [LoadRunner] TruClient Load Test
      >
      >
      >
      >
      >
      >
      >
      >
      > Hi.
      >
      >
      >
      > I am using the TruClient protocol for the first time to conduct a load test
      > and need to have an idea of the resources needed. I downloaded the trial
      > version to my laptop and was only able to run 30 users due to my only having
      > 8GB memory and a 2.5GHz CPU. The load test I need to run requires 500
      > users. The controller I use (also acts as load generator) has 32GB memory
      > and 2 CPU. According to my calculations, I think I will only be able to run
      > 120 users max. Is there any way to limit the memory needed by TruClient?
      > What have been any experiences? We have yet to purchase this protocol and I
      > need to know the costs.
      >
      >
      >
      > Thanks for your help!
      >
      >
      >
      > NS
      >
      >
      >
      >
      >
      >
      >
      >
      >
      >
      >
      >
      >
      >
      > [Non-text portions of this message have been removed]
      >

       




      Reply via web post Reply to sender Reply to group Start a New Topic Messages in this topic (12)
      Recent Activity:
      Yahoo! Groups
      Switch to:
    • Show all 13 messages in this topic