Five minutes. 5 damned minutes. I received a package in my lunch break with a portable hard disk. It was meant as a disk for the virtualized replication target for my main NAS. A replication meant to run twice daily. However, it took me not more than 5 minutes from the “Thank you” for the delivery driver to having a severe trust problem with the newly arrived good. While opening the rack with one hand, the disk fell from the other hand to the ground. Perhaps one meter distance of gravity assisted flight. Directly into the direction of the earths core. Next stop, the stone tiled floor. The disk seems to be okay. SMART didn’t yield any concerning data. I think it has a minimal vibration whil running. But as it dropped before I switched it on the first time I don’t know if this is the way this drive behaves normally. I’m not sure if this is normal. Damned 100€ down the trust drain, essentially.
Now, half a work day later, I thought about two things: At first … I hate my clumsy fingers. My two left hands. But something came into my mind as well. Storage media is the only component of a normal desktop, we must trust. It’s the only component that needs trust. When we drop a main board a few centimeters, or even more, we still use it, if it works, it works. If we drop a keyboard, we don’t think twice, if the keys are still at their place, we us it. If we drop a CPU … well … okay … the “pins” of a CPU are that delicate that even looking at it the wrong way seem to be able to bend the fine spring like contraptions they are using to connect the CPU to the mainboard nowadays. A dropped CPU is probably perhaps not dead but unreachable by the mainboard.
But those components have almost no persistency. Okay, there is a firmware with it’s settting, but that’s easily replicable. We can simply use another one. There is no need of trust. If it isn’t working anymore, we repair it, we replace it. All we have lost is time. And perhaps a few minutes of time with lost quality because of the slight anger arising in us, when the system crashes at the worst possible time. We lose a few minutes of work, thanks to the autosave and recovery functions because we didn’t adhere to the time proven best practice that it’s virtually impossible to persist something to disk too often.
We accept to a very large extend that computers will freeze, network connections will fail, monitors will stay black. But in hard disk we must have trust, we need trust. “Hey, I had this great idea. I have written it/scribbled it down. I made this great photo. I give it to you, hard disk, to keep it safe to infinity and beyond”.
Quite often, the trust has an unwarranted high level. I know by far too many people doing backups - if at all - less often than the occurrence of Christmas and the easter weekend. They never lost data, and then there is the unwarranted assumption, that this will going on this way. There is an infinite amount of trust into the hard disk. As soon as it fails the first time, when the first important piece of data is lost to an accident, the trust drops to 0. And people start to backup. And this trust never really reappears.
There will be always at least this nagging feeling that you forgot something. I’m quite opinionated that you should backup your important data more often than you change your underwear. But I have the impression you many home users of IT must lose their innocence in regard of data loss until they really take this to heart. Having to type an essay for school twice, having to burn the midnight oil to recover a document, that turned into a zero bytes file, prone to a deadline, losing all your flashes of inspiration to a broken promise of everlasting data persistency. A promise that was never made by the device or software to begin with.
That said, perhaps the trust into a hardware device should be always 0. We are talking a lot about zero trust in Security, but we need to think the same about storage, like in never assuming that a single technology will rescue your digitally persisted ideas. I’m probably preaching to the choir, trying to proselytize the already convinced, but looking at the people around me there seems to always some room for missionary fervor.
That said, I have dropped this hard disk and I lost the trust into the drive in an instance. Everything indicates that it’s still fine. But the trust is lost. I’ve now ordered a second hard drive of the same type. I didn’t want to a RAID on the replication target, I just wanted to have a single disk with a replica of the flash pool (one of three, two are local to the NAS server).
Now there will be four. I will keep using the dropped disk. I just don’t trust it any longer to be the only replication target. I will probably replicate to each disk on the remote replication target, using a disk only every 2 days. Given that the disk is running that seldomly the dropped drive will probably have a quite long life despite its sudden stop on the ground. But do i trust this drive? No. Not by a long shot.
On a related note, in the next few days two 16 TB disk will arrive here. I decided not to wait but to invest now. I won’t do RAID on them; it will be purely replica based. The role of RAID shouldn’t be data protection, it should be only availability protection. As many people tend to say, “RAID is not a backup”. I would go further, RAID has no role in data protection. Data protection is just a side effect of this availability potection for certain failure modes, especially for checksumming filesystems. It keeps your data available in case a single or more device fails (depending on the RAID level).
Checksums in the filesystem or replication of data makes protects your data. Backup protects your data, RAID is just helping you to stay away from the tapes, thus reducing the availability impact of an error potentially to zero.
For me it’s not necessary to have uninterrupted availability if a disk fails. I just failover manually to the second disk. But having a time delayed replica protects against a lot more failure modes than having two synchronized disks by the virtue of RAID, because you have synchronized user errors as well for example. One side effect is: There is just one disk running for accesses, the other one is only used for the short replication tasks. Given that a disk is using 6,4 watts while reading/writing, 3,6 watts when idling and 0.9 watts when idling this could be an interesting difference. Of course, when i assume that writing to the second disk for example hourly or daily in one large batch needs less power than doing the changes all the time piecemeal.
One of the reasons why i have chosen such large disk: Many in my family are using system based on MacOS. I will offer them TaafS to the … “Timemachine space as a family service” to them. They can sync their notebooks when they are visiting me to have an encrypted copy of their system. My dad is one of the few Windows bastion in my family and at over 80 years I won’t try to convince him otherwise. TaafS was an idea in a telephone call on Saturday when I discussed with him what I was up to this weekend. I don’t want to explain them again that my knowledge of Windows is rather limited and that I can’t do wonders in regard of lost files. And there my missionary fervor found a new target.
There is a saying “No backup, no sympathy”. Perhaps the way to pity is losing trust.
PS: No, the hard disk in the teaser image on my home page is not the disk of today. It’s a much older disk. This happens to disk I can’t wipe by software. The end of the road for a disk in the house of Moellenkamp can be a very gruesome one.
