3

I have to setup 14TB of raid storage space (which just shows up as /dev/sdb). My question is:

Is there any impact on performance from a kvm guest if I create ext4 on this large drive directly, mount it up on the host like /mnt/kvms and then create raw kvm disk image on it?

Would it be a lot faster to configure /dev/sdb in the kvm machine directly?

I see it as a pro that if something would happen with this large storage area I could not do much with it if I use the device directly but let's say I create 2x7TB kvm raw-disk files on it, if 1 file has a problem at least I still have my data accessible in the other one.

What do you think?

1 Answer 1

4

Is there any impact on performance from a kvm guest if I create ext4 on this large drive directly, mount it up on the host like /mnt/kvms and then create raw kvm disk image on it?

Every time, when your data have more layers to go through performance degradation and latency increase will appear. For example:

  1. Qemu RAW file -> Ext4 -> LVM -> RAID1 device -> phisical device
  2. Qemu raw device -> LVM -> RAID1 device -> phisical device
  3. Qemu raw device -> RAID device -> phisical device

Would it be a lot faster to configure /dev/sdb in the kvm machine directly?

Only you can test it in your workload. With my (fileservers) performance is noticeable better when using (2).

Every time you resign from some layer you lose some flexibility (snapshots, base image, compression etc.), but you'll gain some performance, personally I like (2), because of backup possibilities (less than 2 minutes downtime in a day), size flexibility and balanced performance, using files is also simpler, for example you could mount --bind whole /var/lib/libvirt from backup (to postpone down time) in place of original files in 10 seconds (with one command) to run VM from backup, recreating block devices, editing VM config, is way harder (in terms of time needed).

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .