LTO is definitely worthwhile for the cost, but a bit of a headache too.
LTFS definitely highly recommended. I had used a different software with proprietary format for a few years, but they belonged to same parent company as Drobo which has been in bankruptcy since last summer.
Nothing worse than all your backups kept hostage in a format you cannot read without a software that may not be around much longer.
Transitioned to LTFS and using Hedge Canister as tool. It’s barebones, but has just enough to work which is perfect. Just copying files into job folders and keeping track in a spreadsheet which jobs are on which tape.
1 not happy with softraid, its problematic and quiet frankly… useless.
Thunderblade has a power button called a power cable Ive had many performance issues with the thunderblade as well, t gets hot after long useage and then goes slooow.
I copied 4TB from a internal nvme raid (sonnet 4 banger) to the thunderblade and it took like 4hours… which is really slow.
Really feeling the love for Thunderblade! Bummer I thought it would be a home run paired to Mac Studio. Developing a backup plan now - but since we learn so heavy on Dropbox, really just need to do nightly archives of my timelines to the cloud.
Fast NVMe get very hot and need proper cooling or they will throttle back. I had a long battle over this with two FireCudas. If you just stick them in a slot and want to run 7GB/s it will likely not work out.
And they often sit in tight spaces where it’s hard to put a heat sink on it.
The other issue is how many PCIe lanes you have for them and if any sharing is going on.
You have four lanes going in, and four slots. How are they used? A fast single NVMe when it’s on the motherboard can use 4 lanes on its own. In an enclosure that math is different.
Each TB controller gets 4 lanes on the MB. If serves two ports they get divided up right there. There can be up to 4 lanes going down each TB chain, each device is allowed to takes ones it needs, and passed on the rest. There is some bandwidth allocation if too much happens.
I’m not 100% on this, and it would need testing. But there is the scenario where two daisy chained NVMe enclosures with one stick each could give you better performance than a single enclosure that doesn’t do smart switching and just divides up the lanes in a simple pattern.
In the end TB is still the bottle neck to ultra-fast storage. Which puts Mac at a disadvantage. I’ve not read all the details on the new MP, and while it has PCIe slots, I think it’s lane arrangement is a bit of a kludge and may not be as good as what we know from x86 based boards.
after dealing with my thunderblade again the last 2 days to prep a benchmark for the m2 ultra - having big issues with performance and stability, I cant see myself buying a mac studio,
the storage situation alone leaves me with pretty much no choice than to buy the mac pro instead, nothing I find makes financial sense even if I say i an fine with the tb bottleneck, reliability seems to be hit or miss with all of them except for maybe the iodyne which is too expensive.
So its mac studio with 8tb internal storage or a mac pro with 1-2tb of storage and a sonnet card… price wise not that much difference .
Cant even buy the netstor in europe really, and as far as I can tell its $1000 for each 4x enclosure? so $2000 extra to get speeds that I could get from a single internal nvme drive … i dont know man
also i really dont want to have 20 thunderbolt devices hanging off my computer, I hated that with the mac pro trashcan allready, its just prone to issues having 2 of them … yea no, i dont want hacky solutions to a easy to solve problem,
Just to share my experience, Ive been using a Mac Studio Ultra with the 8TB Thunderblade (Raid 0) backing up to my Synology nightly for just over a year now with zero issues. In the last couple of months I’ve been travelling with it daily, again with no issues. As Andy mentioned I do use the publish method 99% of the time.
I would say that setting up the SoftRaid is a massive pain but I’ve only needed to do this twice (last time to upgrade to Ventura). I had no idea you could use it without SoftRaid so I’m curious about the speeds people are getting???
Literally everyone else on this thread is far more technical than I am so I have no doubt there are short comings to the drive. It suits my purpose which is a mixture of shot based work and tvc so I’m not pushing it particularly hard. I’ve always had a healthy respect for Raid 0 and this thread definitely strikes more fear but for every worst case there’s always a good review so here’s mine!
I just got another email from Support. Basically if a “Softraid volume is not ejected properly, even during a kernel panic, or if the Mac is not shutdown properly” the Softraid volume can fail…
I’m running it raid0 now via apple raid utility. If it’s that fragile in softraid, then any raid scenario in softraid is useless.
Speedwise raid 5 and 0 aren’t that different. about 2400MBps. I assume this is PCI lane technical stuff.
That seems like utter BS. Any good filesystem should be able to recover and rebuild from a power or hardware failure. Especially an external drive by its very design has a high risk of uncontrolled shutdown or disconnect. If that leads to any even reasonable risk of data loss beyond the data that were in transit at the time, the engineers that built that tool should get retrained. That what log replays and other ways of building this are for.
But happy to hear about @drewd’s positive experience. Nothing is ever black and white, and many of these issues lie in the gray space.
And with all that in mind, for any mission critical data you need to follow the 3-2-1 rule, because no single storage is every 100% failsafe, even though it should be better than those Softraids apparently are.