To perform test on ZFS RAID-Z with 4K sector drives

[quentein]

I want to perform test on Free BSD+ZFS with several 4K sector hard drive assembled in different configurations of RAID-Z vdevs. For this i want to access other people system to test performance in several configuration, and I would query anyone to consider contribution his system for testing.
What do you require for testing?
Server_PC which has an net access (port 22, ssh)
64-bit free BSD OS (i advise you for using the Mesa original and licensed CD)
64-bit CPU & different 4K sector hard disk drive which are connected to nonRAID controller which support to FreeBSD.
At least a Ram of 2 GB, extra would be good but not required.
So any one is interested so let me know about this through your reply.

[Mindstorm]

Hey, RAID-Z is a bit odd . It does the some to implement the integrated redundancy which it is same like RAID-3 than RAID-5 really. To avoid the confusion of these whole stuff, I’ll you explain how?
Traditional RAID:

1. In traditional RAID we know stripe-size which is normal of a 128 KB size. rely on the stripe width the “full-stripe-block“ would be “data_disks” *”stripesize” “full-stripe-block“. In RAID-5 the rate of this full-stripe-block is very significant:
2. If we write down exactly the quantity of data of this “full-stripe-block“, the engine of the RAID-5 can do this at extremely high speed, hypothetically the same as RAID0 minus the parity-disks.
3. Suppose we write any additional value that is not a multiple of the full-stripe-block, then we have to to do a leisure read-xor-write procedure which is extremely slow.
4. Traditional RAID-5 engines with write-back basically build up a queue of I/O requirements and scan for full-stripe-blocks which can be written powerfully, and will use slower read-xor-write for any lesser or to spare I/O.

[quentein]

Hey thank for your suggestion and will tell you how does the testing work?

1. We will chat through PM\Email\IRC to discuss all details
2. I attach at the port 22 to your server
3. You can look at with look at utility, while it setup your HDD & carry out testing
4. The testing of the drive may take a few time.

Any (suitable) presentation result I will post in this thread for conversation and advice.

[Eru]

While the other 512-byte sector of hard disk was reading as well as writing at 35 MB/S sustained, this year’s model did not go beyond the 1MB/s barrier.
I make out that for sure that this is associated to the “512-byte” sector firm-ware emulation, because the disk workings rightly well if I partition it in a “4k-sector” alignment. Using it in a ZFS-RAIDZ configuration the performance is very poor because “Raid-Z” uses a “dynamic-stripe-size”.The outcome is that friend like to me, that use several version of Unix, want the firmware to present the HDD as a “4K-sector” disk to set free the full possible of the technology. The OS is by now arranged to support that sector_size, there is no requirement for emulation.