Exhibit A: Hardware RAID5. Exhibit B: Software RAID5. Same controller, same disks. Seriously, fuck hardware RAID.pic.twitter.com/EsIurJpKlQ
Show this thread
And yes this is a cheap controller with no cache, but that's no excuse for that abysmal and not even consistent linear write performance. I'm sure there's some stupid underlying pathological cause for this, but I don't care. Software RAID works without any of this nonsense.
-
-
100% agree... I made the swap to ZFS ~1 year ago, and it's fantastic! What are you using? dm-raid / zfs / etc?
-
Just dm-raid. This is a simple 3-disk RAID5 on a server. I'm slowly implementing Ceph for more serious storage setups.
End of conversation
New conversation -
-
-
Here I have half a mind to say these controllers' firmwarzs are deliberately subpar and here to screw you up if you don't have the expensive cache&battery unit... which vendor will OF COURSE try to peddle as THE solution.
-
I wouldn't be surprised at all.
End of conversation
New conversation -
-
-
Er, these show read rate? And same workload?
-
Write rate. It's literally just cat /dev/zero > /dev/device
End of conversation
New conversation -
-
-
Linux software raid has always treated me well. Probably better then it should.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I was equally surprised when I benchmarked recently, ended up swapping out my RAID-only PERC 6i for a non-raid HBA and let ZFS do all the work.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
What kind of CPU load does that software RAID benchmark generate?
-
2% software RAID, 4% hardware RAID. Don't ask me how that's possible, it just is. I presume driver shittiness vastly overshadows the overhead of software parity calculation (which is so fast on modern CPUs as to be basically free).
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.