He did bit for bit comparisons which do prove the data is intact. If you follow the links through to the plan:
https://blog.za3k.com/flash-media-longevity-testing-4-years-later/ you will see that the whole point was 1) to check for bit rot, and 2) that the drives were usable.
What it proves is that it is common for USB drives to do better than people commonly expect.
The
SHA hashes on the files are missing ( in his cause data blocks so a full image would be required ) so it’s unclear what exactly was done there. Normally, you would generate an
MD5 or SHA checksum at the beginning, then repeat that verification periodically — for example once a year — to prove the data is still intact. I can’t find any evidence of that here, which makes the test rather questionable at best.
In our case, the
SSD endurance test is live-streamed, and every block written is
verified in real time, so there’s no need to rely on occasional checksum comparisons later. Our data is random data in the blocks to prevent advanced controller caching and the block gets written and verified on the same cycle so it would show as a bad block if input and check would be not the same.
Over the past
18 months of testing, I’ve also run into a number of other issues that affect SSD endurance experiments. I have extensive logs on all of this and plan to publish the findings at some point. For now, though, the test we’re running is probably
the closest thing to a real-world SSD endurance test you’ll find online.
No one is calling it a gold standard storage solution. You do not always need storage to be all that reliable: I have micro SD cards in a Raspberry Pi, and in a camera. Fine in those applications as they are not all that write heavy.
Most cameras only use the SD card as a backup anyway. The main recording usually streams to a DVR, so if the card dies it’s not a big deal.
On the Raspberry Pi side we’ve moved away from SD cards almost entirely. We run a few NVIDIA Jetson units in small clusters, and when we tested them on SD cards the performance was pretty terrible. After switching everything over to SSDs the performance improvement was huge.
In general I’ve never found SD cards to be great as OS drives. They’re fine for simple data logging, but beyond that they’re pretty poor in terms of reliability and performance.
What does that prove? No one said USBs are design for sustained write workloads. If the question had been "can I use a USB instead of my SSD" that would be relevant.
We actually tested a range of USB drives, including the supposedly “good” brands. Every single one failed within about 10 days, and most didn’t even reach 100 GB of sustained lifetime writes. It was so bad that it wasn’t even worth publishing the results plus it seem that Power Cycles caused problems too
The real problem with using them for business data is the constant background writes. Applications like Office create temporary files, shadow copies, and autosave versions all the time. That means the drive is quietly writing data in the background far more than people realize.
So when someone suggests storing business data on a USB drive, you have to assume things like Office documents are involved. Those apps generate temp files, shadow copies, and autosaves constantly, which ends up hammering the NAND with hidden writes in the background. On cheap USB flash drives, that kind of workload kills them very quickly.
Depends what you are using it for. How long it sits on shelf can be very important for backup devices, for example. I have USB flash drives with recovery and install images - they might get updated once or twice a month.
My point is simple: USB drives are perfectly fine for things like installers or files that can easily be recreated.
But using them as a backup destination — for example storing
Sage backups when the program closes — is a bad idea. That kind of setup turns a cheap USB stick into a critical storage device, and they simply aren’t designed for that level of reliability.
That is true regardless of where data is. Even if your data in on a RAID 1 array i it should exist somewhere else if it really matters.
I agree that RAID is not a backup solution. However, it can be very useful when maximum performance is required. For example,
RAID 0 ( Dont use if you dont know what you are doing ) can provide extremely high write speeds for certain workloads with very little overhead.
But as I mentioned before, storing business data on a
USB pen drive is about the worst solution you can choose. It’s fine for temporary use or installers, but not for anything critical or long-term.
www.modernretail.co
andrewggibson.com
