Hacker News Reader

Comment by PaulHoule | Hacker News Reader

Those Exos drives have been popular with homelab types because they are usually $50-$100 cheaper than prosumer drives that claim to save maybe 0.5 W and 2-3 db of noise on the datasheet. (The latter could be effaced by a bad bearing anywhere in your build, I haven't seen either claim tested by a review site)

It's funny but in this case the enterprise product is the cheap mainstream product and the 10-20 SKUs aimed at prosumers are the exotic expensive products. It is senseless to me that vendors like WD make special SKUs for security camera use or for 2-3 bay NAS, 4-7 bay NAS, etc when all it means is Best Buy won't stock any of them, development costs are spread out over fewer units, more versions of firmware and hardware to have bugs, etc.

View on HN

> WD make special SKUs for security camera use

The workload for a security camera or a DVR is going to be pretty different from a lot of NAS devices or desktop/server use. Shingled Magnetic Recording (SMR) is fine for things like a security camera, where you're just going to be writing long files and probably just overwrite eventually. SMR is really bad if you need lots of random sparse file modifications.

I do agree they're probably overboard on SKUs but having an obvious SMR SKU versus non-SMR is pretty important in my book. The fact they mixed in SMR into some regular drives and muddied the waters on which tech is going into which SKU really turned me off WD; I don't think I'll ever buy another one of their hard drives again.

Yes

PSA to anyone reading: do not use Purple drives for anything non-DVR/NVR related

Is there something that makes them unsuitable for other workloads?

write amplification. random writes can take, what twice as long?

wear leveling could mitigate, but the choices are somewhat limited:

NILFS seems to make assumptions based on SSDs, blowing up during its cleaning operation

btrfs or bcachefs could help, but who knows whats going on with those

WD mixed both CMR and SMR for the WD Purple, an a few other models too. The CMR wouldn't see any write amplification. But the SMR will suffer from all of the usual SMR limitations. Most importantly that it's only really useful with sequential writes, on top of the write amplification.

Actually the Purple only has one distinguishing characteristic. An ATA command to skip bad sectors instead of trying to read/write repeatedly. This is great for an NVR but it's not mandatory to use.

> WD mixed both CMR and SMR for the WD Purple, an a few other models too.

Could you point me at an example of a SMR-based WD Purple drive? Is this just a historical thing? As mentioned in my other comment, the spec sheets [1] say all (current) drives are CMR.

> Actually the Purple only has one distinguishing characteristic. An ATA command to skip bad sectors instead of trying to read/write repeatedly. This is great for an NVR but it's not mandatory to use.

I'd also be interested to see a reference / details for this.

[1] https://products.wdc.com/library/SpecSheet/ENG/product-brief... https://products.wdc.com/library/SpecSheet/ENG/product-brief...

>WD mixed both CMR and SMR

I was a fan of WD HDDs before and I was in the market for some drives for my new NAS a few years ago. When I checked what drives were on the market, I saw WD doing that while Seagate still clearly marked which drives were which.

I swore off WD that day because my time is too valuable for their bullshit. I have no clue if they have amended their marketing, but I don't care since as far as I'm aware Seagate still clearly marks them and the drives I bought from them have all been fine. I'll likely buy Seagate again next time I'm in the market.

>NILFS seems to make assumptions based on SSDs, blowing up during its cleaning operation

Wait what? Isn't NILFS designed for spinning rust?

The odd thing is that SMR may actually improve performance for many mainstream workloads because it turns random access writes to sequential writes. Similar techniques are used in software for write-heavy workloads (e.g. cache helps with reads more than writes, only helps with writes if they are battery backed and it can help a lot of if writes are sequentialized)

This is why WD figured they could get away with it, but it's a shame what it does if you're rebuilding your ZFS array (if they'd actually talked to customers they might have find a lot of people who buy big drives are ZFS heads!)

Myself I had a security camera set up to monitor my "cat room" last summer when I was trying to gentle a feral cat, I am thinking of pointing it at the driveway and monitoring it with ZoneMinder. ZoneMinder wants its own volume to write to, I am thinking of putting in a spare Intel SSD (superior, though review sites never told you) for that purpose, it's not like I need to store terabytes worth of video from it.

There was an early SMR talk by a WD (IIRC) engineer to a crowd of ZFS devs and sysadmins, and the engineer thought it should fit well because he misheard the default record size being 128MB rather than the actual 128kB, so thought it one record would fit one zone.

That seems like a pretty big mistake, are you sure that's real?

ZFS has a lot of genuine potential to work well with SMR drives, but that potential depends on glue code that nobody has written yet.

I think BTRFS is halfway there?

> ZFS has a lot of genuine potential to work well with SMR drives, but that potential depends on glue code that nobody has written yet.

...and I'm not sure that will change. The thing is, the "drive-managed SMR" models I've managed to get my hands on don't have a zoned storage API. I'm not sure how a filesystem (or application that directly goes to the block device layer) is supposed to work with them intelligently.

There are supposed to be host-managed and hybrid host+drive-managed SMR drives out there, but AFAICT they're only sold to enterprises/hyperscalers (maybe only the latter even). And those are using custom proprietary software (e.g. Google's D servers).

...so who would write/use this glue code? unless the manufacturers decide to contract filesystem developers as a lead-up to wider sale of these things, or something to that effect.

> ...and I'm not sure that will change. The thing is, the "drive-managed SMR" models I've managed to get my hands on don't have a zoned storage API. I'm not sure how a filesystem (or application that directly goes to the block device layer) is supposed to work with them intelligently.

Mostly by making large sequential writes. F2FS is a lot faster on a consumer drive-managed SMR drive than the others (at least until you need to fsck / upgrade kernel version / do any of the actions that utterly suck with f2fs).

Maybe, but I'm not sure I trust that to be enough. The drive's proprietary SMR management firmware might still end up doing something really wasteful under normal usage patterns, in a way that's hard to diagnose and might not be necessary if the application were in control.

Let's imagine an NVR that simply uses the drive as a ring buffer for media data. [1] And it does it via an application using the whole thing at the block device layer, no filesystem. And all the metadata is stored separately on an SSD. This should be the absolute best case for SMR, but even then I'm not confident it will work well. Eventually the drive fills and it starts rewriting. On each write, now it has to guarantee the bytes just beyond your write stay the same. My understanding is it has some minimum write size, or always some fixed-length overwrite, or something to that effect. Using large sequential writes surely helps, yes, and the device presumably has a generous amount of RAM-based write cache, but even so I'm suspicious it'll end up regularly reading and rewriting a bunch of data that will probably be overwritten anyway pretty soon, and even swapping to the CMR area of the drive constantly.

If the application were in control, it'd probably ensure there's some "don't care" buffer between the write zone and the read zone to avoid this. But I don't think there's any way to tell many of these drive-managed SMR HDDs to do that.

[1] You might want a ring per camera stream so you use less read bandwidth on playback of a single stream, and you might want the ability to redact bits of recording in the middle, but let's ignore that.

> now it has to guarantee the bytes just beyond your write stay the same [...] "don't care" buffer between the write zone and the read zone to avoid this.

If the firmware is competent the drive will have TRIM support.

Also it could start overwriting a zone even when there's old live data in it, as long as it buffers a few megabytes to stay at least a track away from the old live data.

And even if it does have to read back the whole zone, it could keep that in RAM to keep the performance impact from being too bad. (Ideally with some kind of power loss protection.)

> If the firmware is competent the drive will have TRIM support.

Sure, if false then false is a true statement.

It's not competent, though? iirc my ST8000DM004 did not support TRIM. I think it's typical of the genre of drives on the market.

I have no faith these drives will not degenerate to the stupidest possible access pattern; their firmware is opaque and bad and their interface is limited.

I know some drives do it though.

I found a comment saying "Sadly it’s 2024 and WD seem to be the only manufacturer actually implementing TRIM."

And I'd expect it to be more common on the higher end drives where someone might intentionally choose SMR to get more space, as opposed to the <=8TB market where SMR is a cost saving they don't want anyone to notice.

> And I'd expect it to be more common on the higher end drives where someone might intentionally choose SMR to get more space, as opposed to the <=8TB market where SMR is a cost saving they don't want anyone to notice.

Yeah, "don't want anyone to notice" is probably a big factor, and in fairness the drive I mentioned was of the surprise SMR scandal era. (8TB was decently large then though I think.)

I guess fundamentally the hidden complexity of the firmware managing the storage also applies to SSD, but I think they've got more wiggle room there due to the inherent better performance of the media, and for whatever reason the average firmware quality appears better IMHO, with some notable exceptions: https://news.ycombinator.com/item?id=32031243

I've tried to find it but it was almost certainly a 10+ year old video and I watched it 3-4 years ago. I do vividly recall screaming at the screen tho, trying in vain to clear up the confusion.

It was a simple mistake, one I've made myself many times. You're so primed to hear A you hear A even though the other party says E.

Anyway, this was of course for host-managed drives, and I haven't heard much talk about that since in ZFS leadership meetings or similar.

ZFS has 16MB block size support, and there has been talk about bumping that up as storage grows, so perhaps one day one could imagine 128-256MB record sizes for storing large blobs, which could fit with SMR better.

But yeah, would need to write code for that and ZFS ain't exactly overflowing with devs.

I believe the latest on zoned devices is that they are scrapping the explicit zone operations API and are letting the OS merely provide a "region hint", to separate writes by the expected lifespan of the data stored. That could be as simple as different zfs datasets using different hints, if one dataset is archival and another contains actively-overwritten data.

BTRFS can handle zoned disks. The doc claims that it was a natural fit with the way BTRFS works.

The problem is that you cannot buy host managed zoned disks.

WD was pushing for this standard a few years back: https://www.westerndigital.com/company/innovation/zoned-stor... I'm not sure what happened to it / whether it was successful or not.

I think they also sold HA-SMR (host aware SMR, a mix of drive managed and host managed) drives to datacenters but I couldn't find any recent news so I guess those aren't that common.

DMSMR is TERRIBLE for a DVR workload. It writes first to a CMR cache and then flushes to SMR when the drive isn't busy. With a DVR the drive is never not going to be busy. Further the pitiful write speed of SMR is likely not enough for a modern high resolution security camera installation.

At least for some brands/models, a workload of purely sequential large writes doesn't exhibit that slowdown -- so I'm assuming they go straight to an open SMR zone. There might be an extra criteria where the rate of writes has be just below some threshold so the drive has time to seal full zone & open up a new one in time, to avoid CMR buffering.

> The workload for a security camera or a DVR is going to be pretty different from a lot of NAS devices or desktop/server use. Shingled Magnetic Recording (SMR) is fine for things like a security camera, where you're just going to be writing long files and probably just overwrite eventually. SMR is really bad if you need lots of random sparse file modifications.

In theory you're right, SMR is not necessarily a problem for NVRs if well-implemented. However, I had poor luck running my own NVR software on drive-managed SMR. (The cheap kind that doesn't support zoned storage APIs, just pretends to be a standard drive.) 2/2 drives failed within a year, despite staying within the rated workload. Admittedly it's a small sample, but it soured me on drive-managed SMR. I suspect NVR software written to use host-managed SMR or hybrid SMR might have better luck, but AFAICT those drives are only sold to enterprises.

...and AFAICT, the WD Purple Surveillance Hard Drive and WD Purple Pro Smart Video Hard Drive lines do not use SMR. Looking at the current highest-capacity drives for both lines:

* The WD Purple Surveillance Hard Drive, 8TB (WD11PURZ) is specced at 180TB/yr.

* The WD Purple Pro Smart Video Hard Drive, 24TB (WD240PURP) is specced at 550TB/yr.

...vs something like 55 TB/yr that I've seen for SMR drives.

https://nascompares.com/answer/list-of-wd-cmr-and-smr-hard-d... also says that WD11PURZ is CMR. No listing for WD240PURP.

edit: oh, and the spec sheets for both lines explicitly say all drives use CMR. https://products.wdc.com/library/SpecSheet/ENG/product-brief... https://products.wdc.com/library/SpecSheet/ENG/product-brief...

SMR is discouraged for DVR use, many vendors (e.g. Ubiquiti) specifically ask for CMR drives.

Or the only difference of those specialized SKUs is a single parameter in the firmware making a different tradeoff for, lets say, random seek vs. sustained sequential reads/writes. A different cache strategy. Or whatever.

Being able to charge $100 more for that is very worthwhile.

I think the "larger NAS" drives have better shock absorption, which is a hardware thing.

No matter what, it's a reason why you can't get any large capacity drive at Best Buy. They'd have room for one SKU but can't afford to stock an Amazonian variety of unusual types. I feel the whole industry has gone in a bad direction.

My workstation has a modern case that doesn't really have a lot of room in it for HDDs, so I added an external HDD for my sports photographs. I sure as hell don't get the bit about "serious external HDDs have their own power supply" because it is one more thing that can fail (get unplugged) and cause data loss. I'd get warmer and fuzzier feelings if the drive was powered off USB.

>No matter what, it's a reason why you can't get any large capacity drive at Best Buy.

Is 24TB big enough for you?

https://www.bestbuy.com/site/wd-easystore-24tb-external-usb-...

Internal, not external. (Sure, people are known to shuck the drive out of the reliability-killing enclosure)

How does the enclosure kill reliability?

External drives are often designed for a shorter duty cycle and don't dissipate heat as well as they could.

I'm mostly concerned about the power cord getting unplugged if, say, somebody is vacuuming around the area. On top of the usual "whole machine gets turned off" failure mode that happens maybe 10-20x a year in my neighborhood because of power loss there is a "machine is turned on but HDD is powered off" case.

Wouldn’t a NAS have the same problem, and wouldn’t a battery backup solve it? Genuinely curious

Personally I don't trust a NAS as far as I can throw it, or rather I'd use one the way I use Amazon S3.

My home server has a ZFS array, the media server and some other programs access it directly. If I want to move files to or from it I use SFTP. If I want to back files up to it I use rsync. I have Lightroom running pretty good on an external HDD, no way I'd take my chances running it on a NAS.

> Personally I don't trust a NAS...

> My home server has a ZFS array, the media server and some other programs access it directly.

This sure sounds like a NAS.

An appliance with the primary role of storage, and the ability to share files over the network are the distinguishing features for a NAS. Network support for iSCSI, SMB, NFS makes a NAS; sharing data exclusively over the media protocols (http, rtsp, etc) makes it a media server

By "NAS" most people mean "box they bought from someone to share files on a network" - think Qnap or Synology. They'll call a "NAS" that is home-built a server, even if they do basically the same thing.

To make it more fun, you'll have people refer to "I don't have a NAS, I run FreeNAS on my server."

There is no SMB or NFS, no 'network filesystem', though I could configure one if I want. I don't believe network filesystems are faat and reliable...

... Now that I think about i did have samba set up so I could watch 3d movies on my Meta Quest 3. But who cares if that is reliable?

People have been battle testing those cheap as possible external drives on top of hot xboxs in enclosed cabinets for years and years now. Backup, and I would think failure rates would be too small to consider.

IIRC the main difference for WD was how it reacts to errors. Consumer drives will spend minutes to retry rw operations. Surveillance drives will mostly ignore errors as it's more important to have some footage than none. And NAS drives will immediately report the errors.

See http://danluu.com/filesystem-errors/ you are damned if you do and damned if you don't

2014-2017 I worked for HGST, just as it was being assimilated by WD.

There were definite hardware differences between drive categories back then, and as well as the firmware adjustments, there would be things such as additional accelerometers to detect and help protect against shock (eg: portable drives), higher quality bearings and head/head mechanism differences.

But, or course, that was then and I can't speak for now.

Surveillance represents a pretty big market though and the OEMs are not getting them from Best Buy.

Most storage workloads read more than they write. Surveillance is the opposite. You are always writing. Much ( maybe almost all ) of what you write is never used and is simply overwritten again.

Similar workload for high energy physics, think CMS or ATLAS detectors at the LHC.

> claim to save maybe 0.5 W and 2-3 db of noise on the datasheet

The huge issue here is that hard drive datasheets have gotten very inaccurate.

They're not wrong, per se, but they tend to make very weak promises that ruin the ability to compare. Like if every car said "more than 10 miles per gallon".

I am skeptical about any claims about noise. I built two computers circa 2001 and read guides on Tom's Hardware and similar sites about how to build a quiet PC and to my schizotypal brain [1] the machines still screamed like banshees and I had to replace HDDs, fans, power supplies and such multiple times until I found parts that really were quiet. My take is there is a high variance in the noise of 'identical' parts, it just takes a bad bearing to produce an annoying squeak 15db or more above spec.

I'm also skeptical about claims about power. Performance loss is real, power savings might not be. I had a $3000 desktop which I couldn't stand to use when I set the minimum CPU low because the machine would go to a lower power state between keystrokes or when I was only moving the mouse around, so it would seem to take forever for characters to appear when I'd type. Granted, my schizotypal nervous system is bad at latency compensation, even when my condition is remission, but I've found trying to save power makes a $3000 desktop feel worse to use than a $300 Chromebook.

[1] particularly inflamed at that time

> make special SKUs

If your company is set up for it, special SKU's can be really cheap. Some can be software-only changes that even get applied after the product gets into the hands of the users.

eg. during the setup process, the hardware phones home, and some server side component matches it up to the original sale at the retailer and figures out which features to enable.

There might even be more SKU combinations than products have ever be produced.

Not saying hard drive manufacturers do it, but it is both possible and done in some industries.

I was impressed in the 1980s when I heard about how 390 mainframes had a modem in them so IBM could phone in orders to unlock extra capacity if the customer needed it for a short time.

Today I think that's a sign to sell the stock of the company involved. It costs money to build hardware that is locked out, either the customer pays for it or the stockholders of the vendor pay for it, if the customer doesn't get to use it it just wasted. A vendor that even considers value subtraction is on thin ice and can only get away with it because there are barriers to exit. (This is particularly true of "fused off" features in Intel chips.)

When you look deep enough, nearly all 'tech' products costs are IP. For IP, it costs nothing to include a feature locked out.

Even real hardware is usually like this. An expensive camera is probably only expensive because it had more R&D go into it and all of its components. The actual sand to make the silicon chips didn't cost any extra.

Unfortunately, the way todays capitalist society works, it's pretty hard to say "I'd like to buy 1 million of your camera sensors please, and I'd like to have 200k of them be high performance and 800k be kinda noisy, but I'll let you know which are the high performance ones after I sell them".

A really vertically integrated company like Apple could do it though - and they do, selling super high performance chips in cheap iPads, but software locked to not run a decent desktop OS - for that you have to buy a macbook with the same chip for triple the money.

The sensors are physically very different in the latest Sony α7, α7R (high resolution) and α7S (high sensitivity) cameras.

In an early minicomputer, the $500 power fail interrupt feature was field installed by clipping a jumper.

There were mainframe line printers that implemented a different lines per minute feature depending on which gear set was installed.

On-demand scaling is still a feature of zSeries mainframes, now called Elastic Capacity on Demand .

To non-corporates it might look like wasted resources or gouging the users, but from the enterprise perspective it's a marvellous feature. It's basically on-premises cloud-scaling.

> the hardware phones home

ah, nothing like Windows-only hardware that is being purchased for Linux usage.

> the hardware phones home, and some server side component matches it up to the original sale at the retailer

laughs in linux

my drives have physical extra pins that make it disable itself with a normal SATA connector that's left unplugged in its plastic case

If a normal SATA power connector disables your drives, that just means your power supply is old.

It wasn't the best idea to reuse the deprecated 3.3v pins for controlling drive power, but it's easy enough to deal with.

Reply on news.ycombinator.com