Notifications
Clear all
SoftRAID Features
5
Posts
2
Users
0
Reactions
1,812
Views
Topic starter
Hi!
Can I get a sanity check on a potential setup I'm considering? I want to make sure I'm understanding any caveats or gotchas, etc.
I have 4 ThunderBay 4's (Thunderbolt 2 versions) and 16x 6TB HDDs. The use case I got them for is no longer needed. My end goal is to have a big storage pool available to my Mac Mini 2018 that I use as a server. I understand the redundancy part of RAID 10 setups but I'm trying to understand how I can maximize performance. Theoretically, RAID 10 should keep scaling up performance with each mirrored pair I add, which is why I'm hoping to add all 4 ThunderBays to the same computer and make one big RAID 10 array. Does that assumption hold in my potential use case of 4 ThunderBay 4's with 16 HDDs on one Mac Mini 2018? Or would anything prevent that? I believe the Mac Mini itself shouldn't be a bottleneck - i7, 6 core, 64GB RAM. It has Thunderbolt 3 and I could plug in the ThunderBays to its 2 independent Thunderbolt channels via Apple's Thunderbolt 3 male to Thunderbolt 2 female adapters and daisy chain the other 2 ThunderBays. I can't imagine even Thunderbolt 2 being a bottleneck, especially if using 2 independent channels. Appreciate any advice on this or if this is a bad idea or if SoftRaid won't work well for that many drives, etc.
Thanks!
Posted : 12/06/2023 11:13 am
4 enclosures should be on two bus's, to avoid disk ejects.
You also would need a "pro" license to do this, FYI.
RAID 10 will scale somewhat, but you need to test this with AJA System test (my recommendation) to see what performance you are getting (use 5K video, 16bit RGBA, 16 or 64GB file size and in settings, select dual DMA engine)
Posted : 12/06/2023 5:36 pm
Topic starter
@softraid-support thanks. I was hoping for some guidance on this before setting all this up. I can't think of any reasons why RAID 10 wouldn't keep scaling up in performance in this situation. Thunderbolt 2 shouldn't be a bottleneck (and I can use 2 independent busses). I doubt my 2018 Mac Mini would be a bottleneck. Can you think of any reasons this setup with ThunderBays wouldn't scale up like a normal RAID 10 would scale up, performance wise? Ie, in SoftRaid or potentially the fact of having 4 enclosures, etc. Thanks!
Posted : 13/06/2023 1:06 pm
Its complex, as disks read ahead. There is not way to predict exactly how all disks behave without testing each. What happens is you need to toss large amounts of prefetched data, slowing performance. that is why it does not always scale.
We ran into this when we tuned the driver for some Toshiba drives a while back, only to find out we would have to tune for all drives. not practical
Posted : 14/06/2023 10:14 am
Topic starter
@softraid-support Ok, thanks. I ended up testing 8 6tb drives in 2 ThunderBays. I think write was a little slower and read was only a little faster. Maybe just 10-20% respectively. Less than I was hoping for. That, the cost of SoftRaid pro, and Apple's unoptimized APFS for spinning drives is making me rethink things. Apparently using direct attached storage with spinning drives in Apple ecosystem is going against the grain. I guess I'm going to have to divvy up my data based on what actually needs to be fast and use SSDs for that. Then use HDDs and RAID only for things that really require the higher capacity where the slower speed isn't a deal-breaker.
Posted : 19/06/2023 6:43 pm
