Performance
Performance
In our 64K sequential read test with a 1GiB test file we see 30.36MiB/sec transfer rate with our 1.8TiB RAID5 array. Running with degraded performance sees a 14.3% drop-off in performance due to the extra overheads. Running the test during the rebuild process sees a further performance drop.
With our sequential write tests, however, we see similar performance in both normal RAID5 and degraded operation, because the difference in performance can be attributed to testing variances. This is where the DS508's 512MiB RAM can act as a cache, hiding the lower performance of a degraded array, and also sustaining higher write than read speeds. Performance during rebuilding takes a significant hit however, running with just 41.1% of the performance of the other modes.
Performance for random reads and writes is understandably lower than sequential ones. Once again, though, we see no performance hit when running a degraded array.
Running with larger 1MiB transfers, instead of 64KiB, we see a more expected level of scaling, with similar drop-offs in performance when moving from a fully operational array to a degraded one, and from degraded operation to rebuilding. The DS508's performance with 1MiB transfer size is lower than for 64KiB across the board, clearly.
