-
Poor write performance on Promise VTrak e610f
My company has a Promise VTrak model e610f with 16 x 2TB SATA drives and a J class expansion chassis with another 16 x 2TB SATA drives. There are two volumes, each with 16 drives configured in RAID 6. This array has been set up for several years and is connected to two Mac Pros and a Windows 2012 server via an 8 Gb/s Qlogic switch. We recently upgraded from MetaSAN to Tiger Store to manage concurrent access. As part of that upgrade, it was necessary to reformat the SAN volumes, so we’re in the process of restoring approx 28 TB of data to the SAN from a NAS and have noticed that the write performance is abysmal at around 30 MB/s.
Read performance is much better (~325 MB/s). For troubleshooting purposes, I connected the VTrak directly to the Windows 2012 system, reformatted the volumes as NTFS and ran some performance tests using iometer. The write speed is OK (~300 MB/s) until the array’s cache fills up (seems to be after 8 GB or so), then it plummets down to 30 MB/s.
With 16 spindles, I’d expect this array to be capable of near wire-speed performance (4 Gb/s) in terms of sequential writes but am not seeing anywhere near that.
Here is a link to a screenshot showing the RAID configuration from the VTrak controller:
https://s13.postimg.org/np0vw0kw7/raid.png
Since we are restoring the data from a NAS with a 1 Gb/s LAN connection, we were expecting the restore of 28 TB to take just over 3 days, which is consistent with the time it took to initially back the data up. However, with the actual transfer rate being approx 30 MB/s, the restore time has increased to over 11 days.
Any suggestions on what I can do to achieve better performance?
Sorry, there were no replies found.