Trust the public cloud Big Three to make non-volatile storage volatile

AWS and Google Cloud virtual machine instances – and as of this month, Azure's – have NVMe flash drive performance, but user be warned: drive contents are wiped when the VMs are killed. NVMe-connected flash drives can be accessed substantially faster than SSDs with SAS or SATA interfaces. The Azure drives – which have been …

[Detective Emil]

Whatever …

Maybe cloud suppliers have successfully conditioned my expectations over the years, but this sort of thing is no more or less than I'd expect. Wonder how fast I can stream in a back-up from dependable storage …

[ecofeco]

I NEVER get tired of saying this

So how's that cloud thing working out for ya?

[Michael Hoffmann]

Re: I NEVER get tired of saying this

Very well, thanks for asking!

Careful and steady as she goes.

Not clear about the purpose of asking in the context of this article specifically, however.

[Thomas 6]

Re: I NEVER get tired of saying this

It's going really well for me and the several multimillion pound companies I've helped transition.

Maybe that's because I actually consider "a high-availability/failover setup" to be a bare minimum in anything that is mission critical.

[Fenton]

Not surprising

Same will happen in any environment where Nvme drives are locally attached.

Same would happen in a VMware environment unless you Vmote the storage.

At some point remote storage in Cloud (and on Prem.) Will move to Nvme.

So personally I don't see the problem.

[Ken Moorhouse]

Will we soon be reading about...

...companies spinning up new instances and finding data left over from a previous session?

[Lennart Sorensen]

Re: Will we soon be reading about...

I believe a single TRIM command to the drive should wipe it, so I doubt they haven't implemented that properly. Issue wipe command before attaching to new instance.

Really not any different than the all other storage for the cloud.

[Ken Moorhouse]

Re: Will we soon be reading about...

I'm wondering about "failure modes" rather than correct operation.

If for example a TRIM command failed for some reason, you would want to shutdown any possible access to that section of the media which failed to be wiped. Is TRIM followed by a "read" check to prove erasure? Obviously there is an overhead in doing this for each section of data, and that is why I question whether this kind of check is considered important, or whether it is glossed over?

[IGnatius T Foobar !]

Tried it and...

I tried it out and my "drive" had 30,000 emails left on it from some previous tenant based out of Chappaqua, NY, USA. I wonder if there's anything interesting in there.

[jaduncan]

Re: Tried it and...

Absent a CVE submission, you should think up more believable lies.

[IanMoore33]

That is pretty FUCKED UP

If I attach a volume ( disk ) .. to an instance I expect it to be valid until I destroy it regardless of the underlying media type

[Anonymous Coward]

NVMe Fabrics on cloud

Actually It is possible to do more with those temporary NVMe drives by setting up NVMe storage server using this software MayaScale from Azure Marketplace itself. It uses NVMe Fabrics over TCP.

https://azuremarketplace.microsoft.com/en-us/marketplace/apps/zettalane_systems-5254599.mayascale-cloud?tab=Overview

More information on https://www.zettalane.com/zettalane-blog.html