Toad World® Forums

Client Machine Compatability


#1

If I have three client machines and all of varying “horse power”, i.e.: dual core, quad core, etc.

  1. Does this cause any problem?

  2. Will Benchmark factory load balance clients across all machines and utilize the processing power on all machines or only use the capacity of the “weekest” client machine?

Thanx,

Ross


#2

This does not cause any problems at all. You can specify the maximum number of virtual users that an Agent can simulate either from the Agent menu under the SETTINGS or via the console by selecting the Agents tab at the bottom of the console to view all the Agents connected to the console and then right clicking and selecting SETTINGS.

Now the number of virtual users that a workstation can simulate before it gets bottlenecked, usually by CPU but can also be memory or network, is dependent on the workload that you are running. Such things as how long it takes for the transaction to execute, the amount of latency, result set size, can affect the number of users. So it is a best practice to run the test first to see how the workload affects the Agent so as to set the optimal maximum number of virtual users per Agent so you can get the most out of your heterogeneous testbed.


#3

If I understand correctly, by default/automatically all clients are assumed to have the same capability but there are manual overrides to allow me to tweak these settings.

If this is true, do you know if there is any intention to automate this process? Not only would this be a signigicant time saver but it would also reduce the risk of knowing all metrics to monitor and intrapret when balancing client workload.


#4

You are correct, by default all agents are treated equally with no maximum virtual user value set, but the settings can be changed as needed. Since there are a large number of variables that can affect the performance of a Virtual User on an agent, there are no current plans to automatically set this value. I agree with the solution offered by kmd, from that test you can see the virtual user count where performance begins to be affected. Of course every test places different demands upon a system so an appropriate Max virtual user count for one test that places low stress on a system may not be appropriate for a high stress test.


#5

Just a thought here - might seem unrelated - but often comes into play.

Best to put BMF agent machines on private network to the database server. Using the public network can run into bandwidth issues. In fact - for hardware you’re talking about, would highly recommend dual giga-bit ethernet cards per BMF agent machien - with the two bound and load balanced as one …