Productive NAS: HW RAID or ZRAID?

NAS group is planned for non-linear editing. There is a question about choose between hardware RAID or sitovym ZRAID. It would seem that the choice in favor of hardware RAID is obvious, but I don't think so. Here's why: hardware RAID takes care of all dispatching functions and optimize the operations of reading record. The same function performs ZRAID, but it uses the main system resources: CPU and memory. In the case when the system is loaded with applications is a strong argument in favor of hardware RAID, but if the system does not execute any application, in addition to the file server, it is not critical. Moreover, in this case (software RAID) to perform the functions of management and optimization of read/write and more resources — CPU and more memory.

Performance interfaces, I / o (i.e. NAS, DAS or SAN) believe it is unimportant, as 10G Ethernet removes the bottleneck at the transmission interface. Total, to build a productive file server use what technology gives the best performance metrics, namely access time to a random file with high number of appeals?

P. S. I Understand that the question a lot of defaults, but it is important to understand what aspects I have not considered in their arguments.
October 3rd 19 at 03:40
2 answers
October 3rd 19 at 03:42
Yeah about the same, especially for such tasks as installation, where a linear reading.
There is greater need to think what to do if the RAID controller dies, or is there a disconnect electricity when writing to the disks.
Oh! It did so, it is important, and that sequence, then it is clearly for one person, but when 5 people are cut from the same array, the question of access time climbs to first place. - arnoldo98 commented on October 3rd 19 at 03:45
Well, even 5 is just 5, not 1000.
Well, or try and use a ssd there is no problem with no response. - Bella.Witting commented on October 3rd 19 at 03:48
any 12 CDs will give you enough for 5 editors. - Bella.Witting commented on October 3rd 19 at 03:51
Now have a NAS (16-2)х2ТБ LSI RAID 6 under Open-E DSS Xeon, 16GB, after a year of use with a significant loss in speed. I sin on fragmentation that should be the place to be. Material may be updated 1TB at a time. Unfortunately, the developer asserts that the store defragmentation is not needed. For me it's not obvious. - arnoldo98 commented on October 3rd 19 at 03:54
use RAID 6 for 12 and 24 screw software for many years, performance did not notice, and all of the drives and give a normal smart car?
The performance of the read or write fell?
What settings the RAID? - Bella.Witting commented on October 3rd 19 at 03:57
Drives whole, the performance drop is defined very subjectively and not by me. Settings immediately-I can't tell you how proetel, never climbed more — downlimit not told. Clients and tasks change frequently, because objective tests are not conducted. Understand that the answer here will not get, but where to dig I can see that. - arnoldo98 commented on October 3rd 19 at 04:00
October 3rd 19 at 03:44
Oh! It did so, it is important, and that sequence, then it is clearly for one person, but when 5 people are cut from the same array, the question of access time climbs to first place.
To remove a response impossible? - arnoldo98 commented on October 3rd 19 at 03:47

Find more questions by tags ZFSRAIDNAS