[glue-wg] GLUE2.1 draft: request for final discussion and approval

[Stephen Burke - UKRI STFC]

Paolo Andreetto <paolo.andreetto at pd.infn.it> said:
> Even better, I think we can remove the suffix "Info" from both the class names. 

That's probably OK, as long as all the names are still different.

> It's not easy to design a model that can be suitable for any kind of future accelerator devices.
> We can have multiple (different) cards on the same WN, and we have some examples about that.

OK, if you think it's possible then you should allow for that - you shouldn't make the model more complicated than it needs to be but it's reasonable to allow for likely future developments. Also you have quite a few accelerator objects, the arguments aren't necessarily the same for all of them. 

> I read about new architecture with multi-chip GPU on the same card.

I think the objects usually only need to be one instance per GPU type, if you have multiple GPUs of the same type you could just publish a count in a single object.

> There're GPU appliances (Nvidia Quadro VCA) that can be shared among different WNs, and I'm not sure that cloning the AcceleratorEnvironment could get the point.

You'd need to have some idea of what the sharing model would be - in general you'd need a new kind of resource object for that, but it may be that you could deal with it within the existing structure. 

Stephen