-
Important Announcement for The TrueNAS Community.
The TrueNAS Community has now been moved. This forum will now become READ-ONLY for historical purposes. Please feel free to join us on the new TrueNAS Community Forums
ZFS Write Performance Issues - FreeNAS 8.0 RELEASE
- Thread starter ChromoX
- Start date
- Status
- Joined
- May 28, 2011
- Messages
- 875
Hi esamett,
I don't think there's much tweaking to be had to help with the limitations of the PCI bus, so I wouldn't expect your situation to improve till you move you move to a more modern controller.
Honestly I think there are a number of things working against you here. First there is the PCI bus speed limit that any one of your drives should be able to max out. Then there are the controllers themselves, neither of which is a particularly "smart" device and each of them require a fair amount of overhead from the OS itself to manage their operation. It wouldn't surprise me to find that the 2 cards are actually stepping on each other with their interrupts as they are passing data across the bus....I think matthewowen01 is on the right track in explaining why reads work better than writes, when the system is reading at least it's in control of the whole process.
I'd be curious what happens if you were to connect just one spare drive to one of the SATA ports and shared it out, I don't think you would see the dips you see now.
Is there any chance you could beg\borrow\steal some more RAM to see what giving the system a bigger cache does to the sawtooth pattern? You could also try experimenting with the "vfs.zfs.txg.synctime" sysctl to see what effect flushing incoming data to the disk more often does...the default is 30 seconds.
I see that you are running a Core2 Duo system so I would expect you have a 16x PCI-e slot. I bet that if you picked up a Dell SAS 6/ir or an IBM BR10i 8 port SAS controller & a PCI video card you would be astounded at the performance you could pull out of your drives.
-Will
I don't think there's much tweaking to be had to help with the limitations of the PCI bus, so I wouldn't expect your situation to improve till you move you move to a more modern controller.
Honestly I think there are a number of things working against you here. First there is the PCI bus speed limit that any one of your drives should be able to max out. Then there are the controllers themselves, neither of which is a particularly "smart" device and each of them require a fair amount of overhead from the OS itself to manage their operation. It wouldn't surprise me to find that the 2 cards are actually stepping on each other with their interrupts as they are passing data across the bus....I think matthewowen01 is on the right track in explaining why reads work better than writes, when the system is reading at least it's in control of the whole process.
I'd be curious what happens if you were to connect just one spare drive to one of the SATA ports and shared it out, I don't think you would see the dips you see now.
Is there any chance you could beg\borrow\steal some more RAM to see what giving the system a bigger cache does to the sawtooth pattern? You could also try experimenting with the "vfs.zfs.txg.synctime" sysctl to see what effect flushing incoming data to the disk more often does...the default is 30 seconds.
I see that you are running a Core2 Duo system so I would expect you have a 16x PCI-e slot. I bet that if you picked up a Dell SAS 6/ir or an IBM BR10i 8 port SAS controller & a PCI video card you would be astounded at the performance you could pull out of your drives.
-Will
- Status
