Test setup

The TS-509 Pro is connected directly to an ASUS Z7S motherboard using a factory made CAT6 crossover cable, configuration is left at default (jumbo frames aren't supported). Local storage consists of four 72 GB 15K disks in RAID 0 on a U320 LSI MegaRAID to prevent any bottlenecks.

To test performance I used ATTO (by mounting a network drive). Additionally files are copied using the Windows copy command.

Performance

My first reaction was "...". I ran the test multiple times, I used an older version of Atto, I checked another host to make sure there wasn't some cache helping out. These results seem valid. Very impressive bandwidth usage.



From the TS509 Pro to my workstation.


And back to the TS-509 Pro.

It seems the TS-509 Pro can burst smaller amounts of data extremely fast. Smaller is relative here though, 350MB files fly. When transferring DVD images speed settles at around 60 MB/s. Still a lot better than most of the competition.
Writing to the TS-509 Pro is a bit slower, though at 40 MB/s sustained there isn't anything to complain about.

Performance under load

When using RAID arrays in desktops people tend to talk about RAID 0. When using RAID for storage, people tend to prefer redundancy. Personally I prefer RAID 5 which allows a single disk to fail without data loss. When replacing the lost disk a lot of XOR calculations have to be done, this can cause quite a hit in performance. To simulate this I created a time vortex aimed at one of the disks removed a single disk, degrading the array. After acting surprised about my array being degraded I put it back in, rebuilding can take hours which should be enough to run a few simple tests.

Performance doesn't noticeably change while rebuilding. This is most likely due to the rebuilding progress being done in the background at low priority. Server class controllers show the same behavior, when priority is set to high the array crawls, when set to low it is hardly noticeable. Rebuilding does take many hours though. This is acceptable, as you shouldn't be doing this often, if at all. And besides, there is no down time.

Power usage

Faster hardware usually means higher power consumption so my expectations were not too great. Power usage peaks at 147W, when idle it's about half that. This is, as expected, higher than the ARM based brothers. However, performance is a lot higher as well.