We have been using up to 5 channels simultaneously over multiple P2 Solo devices (3-4 connected) for a long time. They work extremely well.
We have multiple separate 20gbps USB3 circuits in the machine, the machine has 1tb of RAM, and 120TB of storage plus multiple hot storage pools. We also have 4 GPUs (Ada generation) with zero issues keeping up with realtime basecalling (maximally 40% used).
There is apparently an artificial limitation in MinKnow or my config, or some other strange software block, that is preventing 6+ flowcells running concurrently (without using a virtual machine) on a single running instance of MinKnow. This can be resolved with a virtual machine (but is very inconvenient as we have to partition a GPU to the virtual machine as well as a USB hub at hardware level, and makes maintenance and upgrades difficult).
I would request that the team relax the limitation so that an unlimited number of flowcells run concurrently.
This is not merely a problem of basecalling because even with basecalling disabled, the same behavior occurs.
The behavior occurs randomly in any of the channels -- only 5 will work, and the failing channel plays musical chairs.
All hardware checks pass for all channels and all channels work fine on their own or in configurations of up to 5 concurrently.
We suspect it is some random hard-coded limitation somewhere in a config file that was imposed in the era of the GridIon.
We have been using up to 5 channels simultaneously over multiple P2 Solo devices (3-4 connected) for a long time. They work extremely well.
We have multiple separate 20gbps USB3 circuits in the machine, the machine has 1tb of RAM, and 120TB of storage plus multiple hot storage pools. We also have 4 GPUs (Ada generation) with zero issues keeping up with realtime basecalling (maximally 40% used).
There is apparently an artificial limitation in MinKnow or my config, or some other strange software block, that is preventing 6+ flowcells running concurrently (without using a virtual machine) on a single running instance of MinKnow. This can be resolved with a virtual machine (but is very inconvenient as we have to partition a GPU to the virtual machine as well as a USB hub at hardware level, and makes maintenance and upgrades difficult).
I would request that the team relax the limitation so that an unlimited number of flowcells run concurrently.
This is not merely a problem of basecalling because even with basecalling disabled, the same behavior occurs.
The behavior occurs randomly in any of the channels -- only 5 will work, and the failing channel plays musical chairs.
All hardware checks pass for all channels and all channels work fine on their own or in configurations of up to 5 concurrently.
We suspect it is some random hard-coded limitation somewhere in a config file that was imposed in the era of the GridIon.