Copy 512 bytes/sector disk to 4096 bytes/sector disk

I have recently purchased a new Seagate disk with a physical sector size of 4096 bytes and a size of 1TB. I have other two 1 TB disks, both with a physical sector disk of 512 byte. One of them has FreeBSD installed on it, while the other one has Windows Vista installed. I want to let the new 4096 bytes/sector disk for Windows Vista and use the two 512 bytes/sector disks for FreeBSD so as to play with RAID and ZFS without worrying about sector sizes.

How could I do it right, without degrading performance and respecting alignments? I had thought of dd and playing with ibs and obs, but I am not sure about it and I do not wish to experiment at low level with a new and expensive disk.
 
Don't mess with ibs and obs. Left alone, they will do the right thing, writing 512 bytes for every 512 bytes that are read.

But dd(1) is not the right tool for the job. Use Clonezilla to copy the disk.

After that, there are two ways to go. Use a partition tool to realign the partitions, or assume Seagate's "smartalign" works and just ignore it.
 
In the end I was not bold enough to do the move. I have the Windows Vista system so well acquainted and stable for the last three years and half that I preferred keep it that way. Furthermore, it is on a Samsung disk from 2009, well before Seagate bought it, so its reliability should be good.

What I did was mirroring the two Seagate disks in a ZFS pool, although one was 512 bytes/sector and the other was 4096 bytes/sector. I was able to get an ashift of 12, so no problems from that side.

In normal operation I noticed that writing performance was hindered by the slower drive, which lagged badly behind the faster one, which in turn had many idle times while the other frequently run over 100% busyness as stated by gstat. Reading performance was not so spectacular, since readings where distributed among both disks and adding up, but seldom reaching full potential; it was usual to see both disks working around 50% in sustained reads, and the throughput was halfway between the slower and the faster one.

I ended up detaching the slower disk to dedicate it to other OSes and encryption experiments. I have learnt that there is no much gain by mirroring disks with very different performance, aside safety of data through mirroring and ZFS corrections, but for that purpose it is better to use two disks much more alike, even if they are slow. I am also concerned about the overhead the faster disk puts on the slower in write operations and its impact on reliability.

Note: The area where I have noticed an improvement in performance using my unbalanced mirror is when encryping large files.
 
Back
Top