Have Hard Drives Become Obsolete???

[NookieN]

Originally Posted by ewmayer

If/when the HD fails in will go the SSD, and I will use one remaining HD as the backup for that. In fact I may do the swap sooner that that because waiting ~60 seconds for the HD-based system to come out of hibernate mode, multiple times per day, gets quite wearisome.

I suppose this is off-topic but I'm curious--in what usage scenarios would you put a system into hibernate multiple times per day?

[LaurV]

Originally Posted by NookieN

I suppose this is off-topic but I'm curious--in what usage scenarios would you put a system into hibernate multiple times per day?

On Mercury.

[M344587487]

Hard drives are not obsolete but they are on the way out for the mainstream and long term storage, the niche they still fill is bulk midterm storage. SSDs have come down in price and will eventually beat HDDs in price per capacity which will be the killer blow for that remaining use case. SSDs are also not limited in form factor, M.2 is great for mainstream system builders and intel's ruler form factor is an example of adapting to fit a server's requirements. intel is working on persistent storage from RAM slots (Optane DIMMS) which is interesting. In 10 years it may be the case that CPUs have HBM3/4 integrated into the CPU package with DIMM slots still existing but being for optional additional DDR5/6 working memory and persistent Optane or equivalent storage.

[nomead]

Originally Posted by M344587487

Hard drives are not obsolete but they are on the way out for the mainstream and long term storage, the niche they still fill is bulk midterm storage. .

Maybe I'm not aware of the latest developments in the storage business. What is replacing them in long term storage? Can't be tape, for example, that has only occupied a small niche with very specific requirements for some time now.

[Uncwilly]

Originally Posted by NookieN

I suppose this is off-topic but I'm curious--in what usage scenarios would you put a system into hibernate multiple times per day?

During commutes and other times when it will be closed for 30 minutes or so. With hibernate vs sleep, you can extend the effective battery life (wall time between recharge events) for some schedules.

[CRGreathouse]

Originally Posted by retina

I have banned SSDs from my lairs systems. Too small. Too expensive. And not enough faster to make any meaningful difference. I don't care if a machine boots 10 seconds faster, it's going to be running for months between restarts so I'd much rather have the data capacity (and more money in my pocket) than save 40 seconds each year.

I think it's good to know your needs. I had been running on a tiny, outdated hard drive with maybe 6 full OS installs just sitting there because there was no reason to delete them, so I knew that moving to an SSD was practical for me -- I use very little space, but my speed requirements are high (I do lots of math crunching which overflows RAM). For people with more expansive space requirements, needs will obviously differ.

[aurashift]

Form factors: EDSFF>Intel Ruler>Samsung NF1. They all look the same, EDSFF is the open one. All NVMe flash. M.2 and U.2 both have problems fixed by EDSFF. Its new, but we can fit 800TB in a 1RU server now.



Persistent memory isn't new. NVDIMMs (battery backed DRAM) are a thing, its fast as DIMMs and non-volatile, but you don't get the large capacities like with Optane Persistent Memory. They're best used as cache devices for a storage pipeline, imho.



Optane persistent memory is a 3DNAND chip in a DIMM form factor. (Large capacity with 20GB/s per DIMM channel) Your application needs to support it though.


Traditional spinning rust *may* be good for your warm (not hot) tier. It really sucks in terms of IOPS. Flash (NVMe) is going to get cheaper and eventually displace spinning rust.

[ewmayer]

Originally Posted by Uncwilly

During commutes and other times when it will be closed for 30 minutes or so. With hibernate vs sleep, you can extend the effective battery life (wall time between recharge events) for some schedules.

Exactly. Note that Apple does not actually include a hibernate option - only the still-power-drawing 'sleep' - on Macbooks, so that requires a little hackery to enable, here is what I set when configuring my system to enable hibernate-on-lid-close:

sudo pmset -a hibernatemode 25

[Madpoo]

Originally Posted by aurashift

...
Traditional spinning rust *may* be good for your warm (not hot) tier. It really sucks in terms of IOPS. Flash (NVMe) is going to get cheaper and eventually displace spinning rust.

^^ This ^^

Whether you use SSD or spinning disks depends a lot on your specific needs.

I've had both types fail, recently, at about the same rate. This is on a video recording system, so they get a LOT of write activity.

I keep the most recent recordings on RAID arrays of SSD, so the write performance is up there, and it gives good playback speeds when needed, without affecting all of the other cameras still writing to disk.

For long term archiving, capacity is more important than speed, so a few spinning disks in a RAID array give me the space, and still performs well when doing playback of something older, when I might be rapidly zipping through a day's worth of recording from some camera or another.

Ultimately all the churn will cause failures. I think in my case, my SSD failures to date were due to me using SSD drives that were pulled from machines heading to the great recycle bin in the sky, so who knows how many write cycles they'd already logged (I didn't check). I also try now to move things to the long term archives when there's still about 25% free space on the SSD drives, so I'm keeping a good amount of space available for wear leveling.

The failure of the spinning disks was a mystery to me... they were purchased new, WD purple which is supposed to be good for just this type of thing, but for whatever reason one drive or another fails. I switched to Seagate and we'll see if they hold out any longer.

[retina]

Originally Posted by Madpoo

The failure of the spinning disks was a mystery to me... they were purchased new, WD purple which is supposed to be good for just this type of thing, but for whatever reason one drive or another fails. I switched to Seagate and we'll see if they hold out any longer.

IME the branding makes no real difference. People who have had WD fail on them will then swear to never use them again, and other people who have had Hitachi fail on them will then swear to never use them again, and other people who have had <brand> fail on them will then swear to never use them again, ... etc. etc. etc.

It is just a giant merry-go-round. They all fail eventually. Don't worry about the brand, just buy what you need and plan for failures accordingly.

[Mark Rose]

I just had a 1 TB Seagate HD head crash on me. I have another spare 1 TB drive to restore the RAID, but it got me to thinking: why not just throw the data from that 3-drive RAID5 on a 2 TB SSD and backup to spinning rust? Even 4 TB SATA SSDs are inexpensive now.