Hi,
So for my fileserver I am going with software raid over hardware raid. I want to have the benefit of filesystems like ZFS and BTRFS that have advanced checksumming at the file level that is impossible to do under a hardware raid controller. The other key advantage is that by using BSD or Linux software raid, I’m not forced to use enterprise grade drives that feature TLER, as in software the HDD is allowed to go into deep recovery and the software wont falsely drop it from the array like a HBA card will.
Let’s say I’m exploring the Norinco 20 bay hot swap sas/sata rack mount enclosure for a 20 disk fileserver. What exactly are my options for hooking them all up?
Looking at the detail I’m a bit confused generally about multilane expanders, single lane expanders, SATA port extenders and all the gizmos in this storage space
I don’t want to buy a hardware raid card with say 24 ports, because thats expensive and I don’t need it
Seeing that modern sas and sata drives are on 6 Gb/s bus links, I don’t believe that I need a full lane for each of the 20 disks because the sustained throughput of 7200 rpm drives is well under the 6Gb/s limit. Ok in caching it’s a different story but I’m willing to sacrifice some maximum burst for reduced costs. Most 7200 rpm SATA drives seem to peak around a sustained 120 MB/s or so which is well off from the 600 MB/s that the SATA III bus provdes.
So I was thinking, what if I had a single SATA III 6 Gb/s pcie card with four multilane miniSAS connectors. Then, I could use a backplane extender for each of those four ports, for single lane 5 sata III ports to the one minisas port. So that would give me a total of 20 ports, where from the backplane extender with 5 drives across one SATA III bus, thats a shared 120MB/s of bandwidth allocated to each individual drive.
Is this possible? Can anyone please help with actual part URLs and some guidance on specifics please.
Thanks
So for my fileserver I am going with software raid over hardware raid. I want to have the benefit of filesystems like ZFS and BTRFS that have advanced checksumming at the file level that is impossible to do under a hardware raid controller. The other key advantage is that by using BSD or Linux software raid, I’m not forced to use enterprise grade drives that feature TLER, as in software the HDD is allowed to go into deep recovery and the software wont falsely drop it from the array like a HBA card will.
Let’s say I’m exploring the Norinco 20 bay hot swap sas/sata rack mount enclosure for a 20 disk fileserver. What exactly are my options for hooking them all up?
Looking at the detail I’m a bit confused generally about multilane expanders, single lane expanders, SATA port extenders and all the gizmos in this storage space
I don’t want to buy a hardware raid card with say 24 ports, because thats expensive and I don’t need it
Seeing that modern sas and sata drives are on 6 Gb/s bus links, I don’t believe that I need a full lane for each of the 20 disks because the sustained throughput of 7200 rpm drives is well under the 6Gb/s limit. Ok in caching it’s a different story but I’m willing to sacrifice some maximum burst for reduced costs. Most 7200 rpm SATA drives seem to peak around a sustained 120 MB/s or so which is well off from the 600 MB/s that the SATA III bus provdes.
So I was thinking, what if I had a single SATA III 6 Gb/s pcie card with four multilane miniSAS connectors. Then, I could use a backplane extender for each of those four ports, for single lane 5 sata III ports to the one minisas port. So that would give me a total of 20 ports, where from the backplane extender with 5 drives across one SATA III bus, thats a shared 120MB/s of bandwidth allocated to each individual drive.
Is this possible? Can anyone please help with actual part URLs and some guidance on specifics please.
Thanks
Last edited: