Rob Bradford

How are you accounting for the fact that your RBD device could be accounting for that variability?

I think an implementation of the virtio-block write zeroes feature seems reasonable. The challenge would be integrating it with the block abstraction / io_uring support.

> @rbradford I don't see a challenge with exposing those methods using io_uring. I would add methods for discard/write_zeroes to the [AsyncIO trait](https://github.com/cloud-hypervisor/cloud-hypervisor/blob/97012c511dd5b140793471fbfa27f7fc48288568/block_util/src/async_io.rs#L129) which under the hood would be mapped...

I think you can use generics to have a consistent name and keep the method on trait.

```rust trait SnapshotData { fn as_any(self) -> Any {} } trait Restoreable { fn from_snapshot(T: SnapshotData) -> Self {} } ``` > I think you can use generics to have...

> @rbradford looking at the snapshot code, I was wondering why we ended up with some structures being versionized vs some other structures being serialized? I know we recently moved...

@sboeuf I think improving the quality of the code around snapshot/restore is a great thing do to. Personally I don't think it's a problem to break cross version migration as...

CC @jongwu

Thanks @jongwu - it's possible this is nested (or even nested on Azure) specific - the jump is very obvious in the metrics UI.

> Hi @rbradford -, do you mean that the test is in a nested virtualization environment? Does "cpuid" instruction takes more time in nested virtualization? Yes, our testing runs inside...