Dear cryosparc Team,
we have observed a 10x decrease in performance for cryosparc classification jobs after a gpfs update. Cryosparc and Os are the same.
The problem is hard to debug, as we don’t know how cryosparc is reading the particles.
(seqential read performance is still good 1-2GB/s)
So may question is:
Is there a fio( https://fio.readthedocs.io/en/latest/fio_doc.html ) configuration
that reflects what cryosparc is doing ?
Or is there any other linux tool we can use to measure how cryosparc reads the particles ?Or could you provide us a minimal python code which mimics the reading and locking of cryosparc.
thanks
Florian