[ale] New hard drive procedure

[Michael Trausch]

I don't notice a major difference. Of course performance is lowered for huge files that are Btree COW, but if you set the NOCOW attribute that support is disabled for the tree representing the file. 

I use reflinks and snapshots so I don't disable it.

The performance drop is similar to operations on snapshotted LVM volumes.

The best performance is always offered by a direct unmanaged partition in raw format. Everything else incurs overhead. But as I've stated my VMs all work faster in my current setup due to data being spread over multiple drives which more than makes up for the overhead. 

Everything is a trade off. 

I will see about maybe doing a benchmark this week to show concrete numbers. 

Sent from my iPhone

> On Jun 7, 2015, at 11:51 AM, DJ-Pfulio <DJPfulio at jdpfu.com> wrote:
> 
> http://www.linux-kvm.org/page/Tuning_KVM says
> "Don't use the linux filesystem btrfs on the host for the image files. It will
> result in low IO performance. The kvm guest may even freeze when high IO traffic
> is done on the guest."
> 
> That is a show-stopper as far as I'm concerned for deploying BTRFS.
> 
> Are they wrong?