Hi
I’m currently running some batch jobs on qpu:altair and I’ve noticed some inconsistency in the time that it takes to execute each job.
The batch job consists in running the same circuit fringe_steps times (ranging from 10 to 100), varying one internal phase parameter (from 0 to 2pi):
iteration_list = [ {"circuit_params": settings[i] } for i in range(fringe_steps)]
sampler.add_iteration_list(iteration_list)
counts_list = sampler.sample_count(samples)["results_list"]
The same exact job can take a time ranging from 5 minutes to more than 1 hour ( from the “Start time” cited on Quandela Cloud).
I thus have the following questions:
- Is this observation related to how jobs are scheduled on the QPU?
- Does the batch job take longer because it’s interleaved with some other people’s work?
Concerning the second question, as far as I could observe, longer jobs did not result in worse data.
A side note is that from what I know the phase parameter does not have any major effect on the physical performance of the circuit
A related question I had is if there is any way to obtain information about the job waiting list for the cloud devices, so that one can assess if it’s a good time to submit a job without trying to infer it from the speed of the progress %
Thanks for the consideration 
EDIT: Turns out that I was setting up different photon filters in different Processors which changed the number of samples required.
I solved my problem but I would find anyways interesting to know more about the scheduler