Performance degradation under RAID
I have been using ATIH home to backup my system for years. The backup of approximately 1.5 TB of data on 3 disks to an external drive connected via eSATA takes about 2 hours. I recently upgraded two of the SATA 2 (512 GB each) disks to 2 SATA 600 (2 TB each). The two disks were formely plugged directly into the motherboard; the new disks are plugged into a RocketRAID 640 in RAID 0 mode.
The new configuration, with the same amount of data, now takes more than 6 hours to bakup!!! How does one account for the 300% performance degradation? Is it because:
1) the disks are bigger (although the amount of data is the same)
2) the disks appear as "networked" because of the RocketRAID even though they are still in the computer.
3) because of GPT formatting of the new disks
4) because the new disks were formatted with 1K sector size (recommended by highpoint-tech)
I am running ATIH 2010 with the Plus Pack. Newer versions of ATIH provide additional support for RAID arrays but I have not read anything about improved performance.
Thanks.
I was hoping that someone would already know some of the answers like: a) backup time is dependent only on the amount of data to be transferred not on the size of the file system, b) transfer from "networked" drives is slower by a factor of x, c) 1K sector size is a no no, etc.
But you are right, there is nothing like measurements! Before I present some data here is some relevant information:
a) in the two comparisons (RAID vs no RAID) the system disk (C:) was a Crucial 256 GB SSD connected to the motherboard (no RAID).
b) there was no other activity on the computer
c) the ATIH backup was NOT in block by block copy mode
d) the RAID array was created on the two drives, the disks were formatted and the old data was transferred to the array using MS cut-and-paste from the old disks to the new ones.
I have now disconnected the two 2 TB drives from the RocketRAID array for 2 reasons: 1) ATIH backup took too long and 2) the RAID array would disappear when the system resumed from SLEEP (S3). This is probably a question of timeout between the RocketRAID and the Win 7 OS. I have applied a patch that Microsoft provides to avoid timeouts from large disks but it did not help. I am still interested in the performance question because I want to know if the backup problem would be solved if I got a better H/W RAID controller.
Here are the measurements from ATTO using default values (transfer size from 0.5 to 8192 KB, direct I/O, queue depth = 4)
A. Mesaurement from one of the 2 TB WD drives connected to the motherboard (motherboard supports only SATA 2 so the results should be slower then when the SATA 600 disks were connected in RAID 0 to the RAID array):
READ transfer rates vary from 21 MB/sec for blocks of 512 bytes to greater than 145 MB/sec for blocks greater than 32 KB.
WRITE speeds are very similar but not relevant in this backup example. I do not know what size blocks ATIH uses.
B. Write speeds to the eSATA disk run from 2.9 MB/sec for blocks of 512 bytes to over 108 MB/sec for block sizes > 64 KB. 8K blocks are 40.3 MB/sec.
C. The resulting tib file is 460 GB which means an average backup rate of 32.7 MB/sec.
Correction: the backup in RAID mode took 6 hours, not 4. The average is thus 21.81 MB/sec.
Based on the results of your testing, the data read/write rates are typical values for SATA 2 drives. Did you do any data throughput testing on the RAID0 pair of 2TB drives while on the RocketRAID controller? Did you test the data transfer rate - between - the RAID0 and the eSATA drive?
Concerning your question about backup speed: The speed at which you backup happens will be related to many factors (including settings within ATIH), but the main ones are the data tranfer rates (write) of your target/destination disk and controller, the data transfer rate (read) of the source disk(s)and controller, and the throughput data rate between the two. As long as your source disk read rate is higher than your target disk write rate, and throughput is not severly limited by other factors (system overhead, other apps running, etc.) the limiting factor is the destination disk/controller write speed. Even if you have a great SATA 600 RAID controller running properly on your system, the backup speed will still be limited by the destination disk/controller transfer rate. In your case, I would suspect poor data rate performance on the RocketRAID controller and/or a data transfer bottleneck from the RocketRAID controller through your system bus to the eSATA drive is/was responsible for the extended backup times your were experiencing.