^ Title ^
so I’ve had problems getting linux to actually setup properly but the functional preview on the boot USB stick itself works without issue, so can I just run it that way, or is that going to limit functionality in some way?
^ Title ^
so I’ve had problems getting linux to actually setup properly but the functional preview on the boot USB stick itself works without issue, so can I just run it that way, or is that going to limit functionality in some way?
I’ve run systems directly from USB (installed there, not live distro) and here’s the hiccups I had:
If the thumb drive is just going to be a temporary / rescue system, that’s one thing. I keep several of those in my bag. But for a (semi) permanent install, you’ll probably want to have it installed to a real disk.
Edit: I do have some hardware that boots its OS from a flash drive (Ubuquity router for example) but it’s configured to not make a lot of writes to it and is mostly read only. So for an embedded system, a USB drive could work fine, but for a general purpose workstation, not so much.
There are distros that will boot off a flashdrive, but the whole OS is loaded in RAM so you don’t read/write from the drive. I know a lot of ESXi installs work the same way.
I did this for 7 years and was great. It was for work, we weren’t allowed to format but the boss have me green light for this approach.
Best part is that I could just plug on my desktop and have better performance, and I didn’t need to carry the laptop with me.
It was Mint, then PopOs (with some attempts of fedora).
Yeah, echoing this, I’ve run Linux off of my external nvme enclosure for is testing as well, mitigates the heat and durability issues but was nowhere near the best experience, though it’s supposed to be able to do 10 Gbps so it’s nice in pinch, it’s my rescue and iso drive mainly.
Standard USB keys get toasty as heck just from regular usage, especially the metal bodied ones.
Personally I view SD card installs with a similar level of concern to a USB install, had those crap out with no warning in the past (though tbf, it’s only happened a handful of times), I backup configs for my stuff running off of SBCs for that reason.
Yep. There’s a reason there’s so many Raspberry Pi “boot from USB” how-tos.
Great comment!
Note that @ptz is using bits per second, so divide everything by 8 to get GB.
And, SATA is even slower than NVMe, which is increasingly common, especially in smaller form factors. So, you’re looking at a potential Gen 5 NVMe 128Gbps vs USB’s - let’s say a USBC drive capable of handling Gen 4’s maximum speed of 40Gbps. The number of Gen 4 devices - peripherals or hosts - is rather sparse. Thunderbolt 5 has been announced and will bump bandwidth to 80Gbps, but the earliest possible time we’ll see those is sometime this year, and it’ll be a few years before they’re wide spread and peripherals are common. And if we consider the future, NVMe Gen 6 & 7 have speeds up to 256 and 512Gbps, respectively.
There’s also the consideration that M.2 adoption tends to move faster than USB, so while computers are still being sold with USB 2 ports, and a maybe one or two v3, it’s still pretty rare to find ones that include USB4 despite it being ratified in 2020. But as soon an a new NVMe generation is released, you start seeing more rapid adoption. I don’t know why.
I’m bitter about how slowly the industry has been moving on USB4. For average users, NVMe speeds are already plenty fast - faster is better, of course, but USB3 is still laggy. Gen4, I think, is “fast enough” for most uses; it’s fast enough that most people won’t notice how much slower than M.2 (NMVe or SATA) it is. But it’s so damned rare, and Gen4 docks are still absurdly expensive compared to Gen3 docks.
+1 to all of this.
For ~3 years I ran a Debian system off of a raid 1 of 2 USB drives. I didn’t have the spare drive bay slots in my cs24-ty and I didn’t have the room for an expander.
SanDisk apparently didn’t consider my use case “warranty-voiding” and were content to replace them whenever they failed. (I was honest during the first warranty inquiry about how they were used; I doubt you could get away with this with modern SanDisk though) I had a 3-year warranty on the drives, and checking my email, I replaced a total of 11 over the 3 year period. The first 7-8 were before I moved logging to a zfs dataset on the spinners, which helped a lot as those 7-8 failures were all in year 1 with the constant journaling, writing, and syncing of mostly logs.
TL/DR: great for testing if drivers and hardware work; don’t do this in production
You could install an SSD into an external enclosure and use that as a bootable USB drive.
I did that when I upgraded my M.2 to a larger size.
Can you install software on it? Kernel updates etc.?
I feel it’s the future, keep your data in your pocket, don’t care about hardware.
Yeah, like I replaced my only drive on my Linux install (arch btw) and slapped it in an external enclosure. Then I plugged that in and and booted it on a different pc. You just need access to the bios or boot menu which some public pca might lock you out of.
Yeah, you can. I had to run my file server with the OS on a USB-connected SSD for a few months since I was using all of the internal bays/connectors for the data drives (some of my re-build parts were back-ordered). OP seemed to be implying a thumb drive so I kept my experience to that.