• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

ESXi vs. Solaris VirtualBox Host

H

hotzen

Limp Gawd
Joined
Jan 29, 2011
Messages
349
Hello,

for my upcoming setup I plan to have an ESXi running and hosting only two virtual machines. One being Solaris 11 (Express) with at least 8 GB and the second WHS 2011, so nothing fancy here.
I wonder if it is better to install Solaris 11 bare-metal and virtualizing the WHS with virtual-box?

Would this simplify utilization/passthrough of exotic hardware like tv-cards or usb-hardware?
Is the performance much lower using virtual-box? For a start the WHS would only provide torrent and the nice WHS backup-service, which would actually store the backups on solaris' smb-shared zfs.

If you have ideas for other types of VMs and therefore supporting ESXi, I am open to any suggestions :D
 
Stick with ESXi. I have run VBox on Solaris with a windows VM and it's not stable enough, performance is poor, really poor and it is in no way a contender for esxi. you are on a good thing...stick to it.
 
As someone who went the opposite way, with a Slack bare-metal then a CentOS VBox inside ... I can say I wish I could go back in time and go the esxi route. It really is heads and tails above inOS virtualization.
 
Heh, that was my first go. e.g. I had a centos box with vbox running headless. That lasted about 6 months and I started getting antsy. I went to esxi 4.1 about a year ago and never regretted it.
 
I do have a question - I switched to ESXi 5.0 (60 day trial) on a Xeon E3-1220, coming from native OpenIndiana 151 running Windows XP in VirtualBox 2.4 on a C2D Q9300.

It appears the Xeon machine is about the same speed as the old one, with OI running in ESXi VM. Is this to be expected?

Also, I noticed the CPU usage as reported by VMWare is a bit screwy. OI's System Performance app would report roughly 15% core (4 cores total) versus VMWare's Host CPU usage of 50% per core. Does this mean VMWare has more than double the overhead compared to native?

Thanks for any insight on this - I find that zfs scrubs appear to be slow and they use a LOT of Host CPU so wondering if I am imagining things or not.
 
Are you the guy with non-functioning AES-NI under Solaris regarding encrypted ZFS?
 
Nope - that is someone else - but I read that too hehe!

I am not using encryption. Just 10x2TB Samsung F4EG drives in a RAIDZ2 pool and 8x1TB drives in a RAID0 (stripe). The stripe is much much faster during scrub of course as it is only 40% full (starts at 700 MB/sec in zpool iostat).

Right now my raidz2 scrub is at 30 MB/sec. I must've really fragmented that pool (98% full). :(
 
All filesystems (inkl zfs) gets slow if more Than 90% full. Rule of Thumb: knep utilize under 80% to have it fast again.
 
It's a well known fact that a ZFS should not be filled to rim with data. I'm fairly sure it's even in the ZFS Best Practices, and that utilities like Napp-It has some form of utilization cap enabled by default to prevent this.

If I'm not mistaken, it's related to the copy-on-write sceme that ZFS uses, but someone might correct me on that.

Found it in the Best Practices:

Keep pool space under 80% utilization to maintain pool performance. Currently, pool performance can degrade when a pool is very full and file systems are updated frequently, such as on a busy mail server. Full pools might cause a performance penalty, but no other issues. If the primary workload is immutable files (write once, never remove), then you can keep a pool in the 95-96% utilization range. Keep in mind that even with mostly static content in the 95-96% range, write, read, and resilvering performance might suffer.
 
spazoid said:
IKeep in mind that even with mostly static content in the 95-96% range, write, read, and resilvering performance might suffer.
Click to expand...

The rest makes sense for a copy-on-write filesystem, but why would read be degraded?
 
Yea, I think danswartz is correct. Fragmentation is always going to get worse as you get less available space to move stuff around on the disks.
 
spazoid said:
Yea, I think danswartz is correct. Fragmentation is always going to get worse as you get less available space to move stuff around on the disks.
Click to expand...

But at worst it is only going to effect the read speed of stuff that is written to the "almost full" volume - I still can't see how it's going to effect the read speed of stuff that is already committed to disk - after all it's in a certain location and there aren't any processes that are going to move it - so the read speed should be identical for data that was written at 5% full even when the disk is 95% full - since the read patter for that data would be identical. And just based on the situation virtually all the data is going to have been written before the disk was "almost full".
 
I've done that on an old pool a couple of years ago and I recall it getting "faster". But it didn't affect files that weren't deleted before, as expected.

This time around the slowdown was much more dramatic even though the pool is much (5x) larger with much larger hdds. I guess the older Samsung F1 750 GB hdds must have faster heads than the Samsung F4 2TB drives!

It feels like such a waste that I have to give up 10-20% hdd space to keep speeds up - but then again that's my fault for constant deleting and adding files. If I never delete anything and just keep adding drives, it sholdn't slow down, right? :)

(I am moving files over to another pool temporarily before adding another 8+2 drive vdev to try to balance the fragmentation better).
 
Rectal Prolapse said:
It feels like such a waste that I have to give up 10-20% hdd space to keep speeds up
Click to expand...
Again. ALL filesystems get slow if over 90% utilization. My friend had a Windows PC that slowed to a crawl, and after asking many questions he mentioned that the disk was full. He deleted some files and the PC was quick again.

This is not a flaw in ZFS, all filesystems behave like this. I have a torrent disk and when it gets to ~20GB free (out of 1.5TB) then it slows down. I delete some files and everything is quick again. So I learned to keep more than 20GB free. This applies to my single disk. For a raid the behavior might be different.
 
You must log in or register to reply here.
Back
Top