HDD for cctv

I have been pursuing 2 x warranty claims against WD (now SanDisk as far as I can see). The instructions on their website were incorrect but I managed to get hold of customer services who were helpful but slow. The claim started in January, one has now been replaced, I sent a reminder earlier today that I am still awaiting the other (they apparently have a stock shortage).

I bought 4 x WD Blue 2 TB SSD in 2020, these two have failed, I expect the other 2 to fail just after the warranty expires.

I have 4 x 4TB Seagate Iron Wolf spinning drives in my backup process bought in July 2024 so too early to comment on reliability otherwise I would stick to Samsung, not had any of theirs die on me. Crucial have been good too, I have had one die but their MX series, better than BX, is now discontinued.

The only real answer is real time backups, spinning disks will usually give you some warning of their impending demise, my failed SSDs died suddenly and catastrophically with no warning.

Just a comment, drives specifically for DVRs and CCTV do exist, though of course cost a bit more.

As far as PC drives go, I've seen proportionally more failures from WD that others, and in the very early days of hard drive video recording, Quantel used to use Fujitsu drives exclusively for Harry. Of course, that was decades ago, I've had some Fujitsu PC drives but I don't know if they're still in that market.

You'd probably do better trying to find a security forum, where there should be plenty of appropriate experience.

But that's what a WD Purple is for, is multiple-CCTV-threads recording.

The WD Blue is a terrible choice, for any purpose, as it has the most-crude voice coil control (for a relatively high density platter). As far as I know, it would have an eight hour usage day, 110TB/year, as a metric for wear characteristic. I don't think I have a spec handy for it. If you "had to buy cheap", I'd tell you to get a WD Black instead, as they're one level above the Norwegian Blue. I received a WD Blue in a refurb computer and... had to replace it. There is a WD Black there now. The drive was by itself, so it wasn't a vibration problem, which is what most people would accuse you of.

The Purple should have been a bit better. It's 180TB/year so has a higher rating than the Norwegian Blue. 3 year warranty. The WD Purple 3TB (or larger) has 256MB cache. The lower capacity ones have 64MB cache. Modern drives, the cache really works, and with NCQ tagging, the drive can reorder writes to best suit the seek pattern. The computing device writing to the drive, has to have AHCI to have the driver support for tagging of commands. (SAS drives have really deep tagging, the SATA tagging is more of a joke.)

The WD Gold is 550TB/year and is considered a "continuous duty" drive.

24 hours a day. It has multiple levels of actuator. A voice coil for gross control. A piezo located near the head for fine control, and a DSP and NRRO control to help null out the effects of vibrations from adjacent drives. The WD142KRYZ is 14TB, helium filled (matters not for wear characteristic), has a 512MB cache (bigger than a purple, but not multi-thread optimized). It could still be multi-thread optimized for server operation though. It's a 512e drive (your file system should be aligned properly, not "formatted on WinXP" which would be a mistake). The 512e drives do not do as well if MSDOS formatted by older OSes with divisible-by-63 numbers. And items like that (Enterprise server drives) make a "thumping sound" during self test, which is quite frequent. They don't go whole-hog and make the screeching sound that 15K RPM drives made every 71 seconds, so they're not that weird. But they do still thump. Warranty is five years.

Drives bigger than the 14TB, simply have more platters inside. The platters on those are relatively thin, and the base material is likely different (back to glass platters). There isn't much benefit going to the 26TB model, except it has OptiNAND (writes the 512MB cache to a NAND flash memory on emergency power fail, copies the NAND to the disk when power is next restored). That allows the heads to be parked immediately, without attempting to squeeze in any last-second writes.

A rule of thumb, is drives in web servers, constantly doing random seeks, the arm lasts about one year doing that. A CCTV recorder could be moving constantly, but perhaps at a slightly lower rate-of-seek (as the cache can queue up the stuff a bit for write). The multi-channel recorder can also help out, by increasing the size of segments recorded for each camera, so the seeks per second are reduced a bit.

The Purple drive in this case, should really last more than a year.

*******

And a warning about the drives mentioned so far. SMR utilize writes in 7 track chunks, and would be TERRIBLE for a CCTV recorder. The poor drive would be saying stuff like "I'm drowning here, cut that out". SMR drives are only suited as backup image drives (continuous write of a single 1TB file kind of thing). They would not like handling small files. They're not NAS material for example, Not particularly suited to RAID arrays. SMR writes of small quantities of data, use Read-Modify-Write of 7 track chunks. You can see in this WD Blue document, there are a couple turds in the punch bowl, to be avoided (like, if there was a reason to be buying a WD Blue in the first place).

8TB 8TB 6TB 6TB 4TB 4TB

WD80EAAZ WD80EAZZ WD60EZAX WD60EZAZ WD40EZAX WD40EZAZ

CMR CMR CMR SMR CMR SMR ^^^ ^^^ Lemon Lemon

There was a stink, when some WD Red drives intended for NAS usage, were discovered to be SMR inside (which won't handle write updates well at all). Now, the parameter is more likely to be listed in a datasheet, making it easier for customers to avoid such workmanship.

# Lemon, do not reward the industry by buying these.

# perpendicular magnetic recording, PMR # == conventional magnetic recording CMR # This is the good kind. Bits stored vertically. 0.5 to 26TB currently.

# HAMR \___ high capacity, expensive # MAMR / Not likely for home use (yet)

*******

Make sure you're aligning the file system properly on these drives. If the recorder is "an ancient piece of crap", it could be partially responsible for the bad behavior. Win7..Win11 have

1MB alignment and power-of-two alignment which works well with 512e (4K internal sector) hard drives. Drives with 512n sectors, don't care about alignment, but the biggest one today is a 2TB 512n.

I have one drive here, which is a 4TB 512n, and that might well be the highest capacity 512n drive made. The sectors internally are 512 bytes. The sectors externally are 512 bytes. There are no tricks or read-modify-write involved with such drives, when the file system is not aligned properly.

Properly supporting 512e drives, when they were invented, is why we have real working cache designs in modern hard drives, and why they can accept 1200-1400 commands per second (even though the platter rotates 120 times per second). The drives in the old days with 2MB or 8MB cache, those were mostly a bad joke -- the RAM was at least partially used for the spare sectors table.

512e versus 512n (there aren't many 512n models left)

sudo fdisk /dev/sda

Welcome to fdisk (util-linux 2.37.2). Command (m for help): p

Disk /dev/sda: 3.64 TiB, 4000787030016 bytes, 7814037168 sectors Disk model: ST4000DM000-2AE1 Units: sectors of 1 * 512 = 512 bytes Sector size (logical/physical): 512 bytes / 4096 bytes <=== 512e drive, align partitions to power-of-two for best performance I/O size (minimum/optimal): 4096 bytes / 4096 bytes Disklabel type: gpt Disk identifier: FD3FB597-4327-44E8-9D4B-DD88747391CA

Device Start End Sectors Size Type /dev/sda1 2048 1026047 1024000 500M EFI System

Command (m for help): q

mint@mint:~$ sudo fdisk /dev/sda

Welcome to fdisk (util-linux 2.37.2). Command (m for help): p

Disk /dev/sda: 3.64 TiB, 4000787030016 bytes, 7814037168 sectors Disk model: WDC WD4002FYYZ-0 Units: sectors of 1 * 512 = 512 bytes Sector size (logical/physical): 512 bytes / 512 bytes <=== 512n drive, no alignment needed, "the way they USED to make drives" I/O size (minimum/optimal): 512 bytes / 512 bytes Disklabel type: gpt Disk identifier: CB193B17-CFCE-4E6B-A467-8627288B88B4

Device Start End Sectors Size Type /dev/sda1 2048 206847 204800 100M EFI System

Command (m for help): q

*******

These are what the industry tells you to use. Whether the more fancy drives are better in the long run, who knows. The caches on these, are expecting write threads to be pointing all over the disk, instead of the writes being "intelligently positioned". You would want the largest cache possible on a disk like this, for best longevity.

The fake drive label on these, indicate their intended application. The real labels as shipped, have serial number and so on.

The fake label on the Seagate, is more fluffy. This is their CCTV disk. It basically matches their competitor.

Since the RPMs are not listed, this is likely a 5400RPM drive (consistent with a 5.3W power consumption).

This is the Purple. It also does not list RPM, so I have to assume 5400RPM as well.

Disk RPMs

15000 Server disks 300MB/sec transfer rate (you cannot sleep in the next room, with these!) 10000 Velociraptor/extinct. 7200 Average home computer boot drive (WD Black) -- "real men use an SSD" 5400 Low power drives (backup drives can be this way) 5200 3600 The tat we used to buy :-) Just awful. [Marketing People. Could we do without them ?]

The existence of this one, tells you there is nothing that special about the transport. Notice it is 7200 RPM, unlike the others, and has a 512MB cache, and the firmware discipline is multi-thread cache for video. 6.3W, a bit more power usage.

550TB/yr, like a WD Gold of the same capacity.

WD Purple Pro WD181PURP 18TB 7200 RPM 512MB Cache SATA 6.0Gb/s 3.5" Internal Hard Drive

Yes, looks like Seagate has a response to that one. Maybe an Ironwolf Pro in sheeps frock. Half the cache.

Seagate SkyHawk 3.5" 16TB HDD SATA Internal Hard Drive 7200RPM 256MB Cache

Summary: No one can afford to buy one, but the WD181PURP has some sweet specs. I own 30 hard drives, but I could only afford to buy 1 Big One :-) That's a Helium, to see how long the gas stays in them :-)

Paul

That made Gemini AI tie itself in knots ...

"A hard drive spinning at 15,000 RPM (Revolutions Per Minute) would make one complete revolution in 71 seconds (approximately). Explanation:

15,000 RPM means that the hard drive spins 15,000 times in one minute.

There are 60 seconds in a minute.

Therefore, to find the time for one revolution, divide 60 seconds by the RPM: 60 seconds / 15,000 RPM = 0.004 seconds per revolution.

To convert this to seconds per revolution, multiply by 1: 0.004 seconds/revolution * 1 = 71 seconds/revolution"

Drives for PVRs and CCTV do not have to be high performance and so you can use disks with lower spin speeds (5400 rpm or lower) and longer read write times. These disks usually have a much lower power consumption and run a lot cooler. The cost is normally lower for such disks.

Which is just better than half the price of a half-decent SSD.

Would the SSDs outlive the HDD?

Yes.

First we get a workload figure.

"On our system with eight cameras recording continuously, we’re writing 96 TB of data to the drive each year. This puts us well under the 180 TB/year annualized workload rating [of a small WD Purple not Purple Pro] – meaning the drive should last for a long time." ~7 years in his example, you're lucky to get 50,000 hours on a drive, which is 5.7 years

96 TB <=== This will be our workload estimate per year, for illustration. 96,000 GB 96,000,000 MB

365x24x3600 seconds 31,536,000, which is 3MB/sec for 8 cameras You might want to compare that to what is known about the CCTV setup. Those must be something like 640x480 cameras, as that figure seems a bit low. Maybe the video compression format uses a long GOP.

CT4000BX500SSD1 4TB capacity 1000TBW <=== 10 years of recording at 100TB per year

Samsung 870EVO 4TB capacity 2400TBW <=== 24 years of recording at 100TB per year

Since a hard drive might last 50000 hours (with modern drives, we don't have a good way of predicting this!), which is 5.7 years, the SSD is looking pretty inviting.

I had a disk that lasted seven years at work, and it might not have had an FDB motor in it (which is frictionless). If you used a Helium drive, those are unlikely to have a fate determined by contamination, but then again, the gas isn't going to stay in the thing for 20 years. An air filled drive with a "breather hole", some effects are apparent from what gets inside the drive from the outside.

The drive I have with 5.7 years plus on it, the drive is still in good shape. The computer it was connected to, it died, and the drive has not been used since. Being the drive has an FDB motor, and it never parks the heads, that drive may end up being a record setter, if I cared to continue testing it. I have no way of predicting 20 year lifespans, at the moment, the evidence isn't there.

OK, you win, SSD it is. It might be a bit cheaper, and have fewer caveats. While an SSD could disappear after a power failure, mine has been through a few power failures with no evidence of harm. There have been anecdotal reports from someone who repairs computers for a living, that the brand I use *has* had fallout, and drives returned under warranty. Must be my lucky rabbits foot.

Now, just make sure the 96TB of data per year, mentioned in that article, isn't far off the real device.

Paul

It is getting to the stage where they will.

Especially well oversized ones

Very good video by Christopher Barnatt on that subject..

Old fairly slow one bit SSDs use less power and are more robust.

I had one fail, but not from old age. It was fairly new. The rest have been going a long time and still show no errors. If you care to look using SMART data you can tell if they are 'wearing out'. In plenty of time to replace.

One way to determine how strenuous the workload is, is to put your hand on the drive, while the recorder is running, and "see what the thump rate is". If the drive is rattling like a pig, then for whatever reason, the workload is too much for it. If the application caching is good, then the thump rate should be low or not apparent.

You still have to figure out why a Purple failed after 2 years. Something isn't entirely right there. Overheated ? One of those recorders without a decent sized fan ? An FDB motor does not have to like heat -- you can blow lube past the seal. It could be a bit happier if kept at a moderate temperature rise. FDB motors are good, but they're not magical. They have to live their entire lives on *2 drops* of oil.

One of the disk drive companies, tested those third-party motors in a reliability lab. And to detect end of life, they used a milligram balance, to work out how much of the two drops of oil were left. When an FDB motor dies, it seizes. Someone in a newsgroup reported an FDB based drive inside an enclosure and sitting on the floor, it did a "little hop" when the platter stopped spinning instantly. The enclosure was seen to move.

Much time has passed, since the FDB motors behaved that way. I think the design is proven now, with no recent reports about them. They can be "secured" at one point or two points. The third party company, made some motors which are only secured on one end. The result of this, is the motor makes a rumbling sound during spin-down. Since the year they tried that cheap s*it, they are now (by observation), back to being secured on both ends. The "securing" refers to keeping a bearing surface present for the spindle until the oil pressure comes up and the spindle enters the frictionless state. But there is a potential interval at startup and shutdown, where friction is present. the quieter the drive is, the better it is for you :-)

Paul

Except in the case of one of my WD Blues where SMART data returned zero for everything! I think that was well defective.

I have been using WD Red (Pro these days) for some years. I have had a couple of warranty returns, but mainly they have been very good. I currently have a number in service. Seven of them have 75000 to 85000 hours logged (i.e. nearing 10 years). This is in continuous use.

WD = HDD Sandisk = SSD

Wheras I used to get one marketing email most days, I now get two.

Helium is the Houdini of all gases, and will escape from pretty much anywhere. It is used for leak detection in vacuum systems for exactly that reason.

When they started using it, they must have known that sooner or later, it would be in short supply.

The thing is, we could get more of it, if we wanted. Some natural gas wells have enough Helium mixed with them, to make it worthwhile separating the Helium. And as the supply tightens, that should increase the interest in extracting it.

*******

The disk drive design is weird, as the cover has an adhesive around the outer edge, and the claim was, the adhesive area was large enough along the direction of potential flow, to limit the escape of the gas. Yet, when you look at video of a helium drive being disassembled, the adhesive application section doesn't look as generous as the blog entry let on. I don't know why an adhesive would control the escape, when so many other materials don't work.

Paul

The US government considered helium to be a strategic resource and stockpiled it. When they changed their minds they sold off their reserves and that made helium artificially cheap for a while.

We know it's there and we know how to extract it. But while the US is selling off their reserves it's not economic to do that, This has been worrying some people for a while. Helium is the coolant for the magnets in MRI scanners and it's starting to get more expensive.

The adhesive is sandwiched between metal surfaces. The helium has to cross the width of the advesive blob, not its thickness. Helium will diffuse through any material but through some faster than others.

The eqations that estimate the rate of diffusion are Fick's law and Graham's law.

The hospitals seem to vent MRI boil-off to atmosphere, and don't typically have a compressor.

The physics lab in university, someone did some forward thinking, and even when the gas was cheap, there was a four stage compressor in the basement and a large collection "balloon". And rooms in the building would have hot and cold water taps, plus a third tap labeled "Helium return", and that's where you would connect your cryostat. You could book out 100 liters, and most of it would get re-liquified and available for someone else. This made some improvement in helium expense, and likely now more than ever.

I don't get the impression hospitals do that. Maybe some day, they will. (They have to be able to vent quickly on a magnet quench, but otherwise for low rate boil-off, they could be recycling.)

Paul