Subject: Re: [Patch v5 0/3] Introduce a driver to support host accelerated access
to Microsoft Azure Blob for Azure VM

On Mon, Oct 11, 2021 at 05:46:48PM +0000, Long Li wrote:

quoted

quoted

Subject: Re: [Patch v5 0/3] Introduce a driver to support host
accelerated access to Microsoft Azure Blob for Azure VM

Greg Kroah-Hartman [off-list ref] writes:

...

quoted

Not to mention the whole crazy idea of "let's implement our REST
api that used to go over a network connection over an ioctl instead!"
That's the main problem that you need to push back on here.

What is forcing you to put all of this into the kernel in the
first place?  What's wrong with the userspace network
connection/protocol that you have today?

Does this mean that we now have to implement all REST apis that
people dream up as ioctl interfaces over a hyperv transport?  That
would be insane.

As far as I understand, the purpose of the driver is to replace a "slow"
network connection to API endpoint with a "fast" transport over
Vmbus. So what if instead of implementing this new driver we just
use Hyper-V Vsock and move API endpoint to the host?

Hi Vitaly,

We looked at Hyper-V Vsock when designing this driver. The problem is that

the Hyper-V device model of Vsock can't support the data throughput and scale
needed for Blobs. Vsock is mostly used for management tasks.

quoted

The usage of Blob in Azure justifies an dedicated VMBUS channel (and sub-

channels) for a new VSP/VSC driver.

Why not just fix the vsock code to handle data better?  That way all users of it
would benefit.

Hi Greg,

The Vsock is connection based.  On both Guest and Host, the model is around a connection over a socket. Internally, the Hyper-V creates a device for each connection. But it doesn't scale to large number of CPUs over a socket connection. Hyper-V vsock is designed to perform configuration management for a VM running on Hyper-V.

The Azure Blob service is not connection based. It doesn't fit into the device model presented by Vsock from Hyper-V.

Thanks,

Long

thanks,

greg k-h