When the distribution policy is set to global, is it possible that virtual users could potentially use the same value from the dataset?
1. Does the likelihood of this happening increase if you have a greater number of load generators and virtual users?
2. Is it more likely to be two virtual users on the same load generator using the same value, or two virtual users on different load generators?
A variable value distribution has three available settings
The value scope can be:
Local: The values are specific to each Virtual User.
Global: The values are shared among Virtual Users.
Unique: Identical to Global, with the guarantee that a value is only used by a single Virtual User at a time.
Values may be distributed in a set order:
Sequential: Values are taken in the order they are defined.
Random: Values are taken in random order.
Any: Best effort sequential order. This order is more efficient than sequential order.
When no values are left
Several policies can be applied:
Cycle: The values are used over again, starting from the top.
Stop test: The load test run is stopped when all the variable values have been used.
Return the value "": All subsequent requests for values return this value.
So, if you want to ensure that it is less likely any values are shared, you would use "Unique" for Scope, and "Any" for order. The use of shared values between two or more virtual users depends on if you have enough values total. And, if you run out of total Unique values (When no values are left), what do you want to happen (See above). This info is from the following document:
Hi Keith, You have told me the best policy to make it less likely you are sharing values, but you didn't actually answer any of my questions.
Could you please deal with each question in turn and provide an answer? Thanks in advance,
I think Keith has given a perfect solution. However, let me try answering your questions directly:
1. Global means it will be shared by each VU. So each VU will share the data. Increasing or decreasing the number of LGs or VUs will not affect the likelihood in such case.
2. Again, each VU will start from the 1st point of the data. So the likelihood remains same in both the cases.
Hope this helps.