Good hard drive for streaming lossless??

[somebodyelse]

You're talking about conventional SMR drives again, and missing the point. They aren't a problem - the OS can detect them and schedule writes to suit them, and they have consistent behaviour. The problem with these WD Red models is that they try to appear to the OS (and until recently to the customer) as though they're ordinary drives - they're calling drive-managed SMR I think. This works fine until you continuously write more data than they've allowed for in their CMR buffer, at which point it stalls for an extended period while it moves data from the CMR buffer area to the slower SMR area. We've already covered the rebuild situation when you're adding one of these drives to an array. The other problem area is if you've created an array with these drives, then you write too big a block of data in one go. Now the stalls can happen on more than one drive, so you can have multiple drives marked as failed, potentially taking down the whole array.

 

[Berwhale]

the OS can detect them and schedule writes to suit them,

I didn't know that, that's interesting, but apparently not all OSes support SMR in this way (happy for someone to correct Wikipedia)...

A number of file systems in Linux are or can be tuned for SMR drives:[27]

• F2FS, originally designed for flash media, has a Zoned Block Device (ZBD) mode. It can be used on host-managed drives with conventional zones for metadata.
• Btrfs ZBD support is in progress, but it already writes mostly sequentially due to the CoW nature.
• ext4 can be experimentally tuned to write more sequentially. Ted Ts'o and Abutalib Aghayev gave a talk in 2017 on their ext4-lazy. Seagate also has a more radical "SMRFFS" extension from 2015 that makes use of the ZBC/ZAC commands.[28]
• For other filesystems, the Linux device mapper has a dm-zoned target that maps a host-managed drive into a random-writable drive. Linux kernel since 4.10 can perform this task without dm.[29] A zonefs from 2019 exposes the zones as files for easier access.[30]

In addition to Linux, FreeBSD also has protocol-level support for host-managed SMR drives.[17] As of April 2020, neither Windows nor macOS supports the ZBC/ZAC commands required for such drives to work.

Source: https://en.wikipedia.org/wiki/Shingled_magnetic_recording

 

[somebodyelse]

When I said OS I was assuming linux and FreeBSD as between them they cover most NAS options. The lack of normal detection is covered in some detail in the smartmontools bug report, as are some of the failures with RAID, and WD's dishonesty when asked about what was going on. The pattern of initial denial, followed by blaming the customer for not knowing about the undocumented 'feature' is very much like an issue with a power saving 'feature' on some Blue and Green models I ran into a few years ago. It couldn't be turned off, and IIRC with default linux filesystem options it resulted in frequent head reload cycles and subsequent premature drive failure.

 

[ZolaIII]

I've heard the firmware in those video surveillance drives will skip writes on sectors its having trouble with in order to keep overall throughput up and continue writing footage to the drive.

Skipping some frames here and there may be a good tradeoff for surveillance, but not for general file storage.

That whose more like a guide line, not set in stone. Cheap, fast enough for the purpose even optimised for ti, relatively silent for a mechanic drive and long lasting. That's called SMART and its tuned even more on such product to keep them running. Nothing new really & nothing out of ordinary.

 

[tifune]

Those figures should be taken with a pinch of salt. They use cheap consumer drives well outside their intended parameters.

Can you elaborate on that a little?

 

[mansr]

You're talking about conventional SMR drives again, and missing the point. They aren't a problem - the OS can detect them and schedule writes to suit them, and they have consistent behaviour. The problem with these WD Red models is that they try to appear to the OS (and until recently to the customer) as though they're ordinary drives - they're calling drive-managed SMR I think. This works fine until you continuously write more data than they've allowed for in their CMR buffer, at which point it stalls for an extended period while it moves data from the CMR buffer area to the slower SMR area. We've already covered the rebuild situation when you're adding one of these drives to an array. The other problem area is if you've created an array with these drives, then you write too big a block of data in one go. Now the stalls can happen on more than one drive, so you can have multiple drives marked as failed, potentially taking down the whole array.

I don't see how a drive with such behaviour can possibly be in compliance with the SATA spec.

 

[mansr]

Can you elaborate on that a little?

To keep costs down, Backblaze buy cheap consumer drives, sometimes even scavenging from USB enclosures, and mount them in custom-made cages where they experience temperatures and vibration far in excess of what is normal for a desktop system. Those drive cages have gone through several iterations with variously effective cooling and vibration damping. They've written about this on their blogs. Results from that environment simply do not transfer to domestic use.

 

[Berwhale]

I don't see how a drive with such behaviour can possibly be in compliance with the SATA spec.

What part of the SATA spec would the drive not be in compliance with?

 

[mansr]

What part of the SATA spec would the drive not be in compliance with?

Are drives allowed to take an unlimited time to complete a command or return an error?

 

[Berwhale]

Are drives allowed to take an unlimited time to complete a command or return an error?

Did you mean ATA/ATAPI when you said SATA?

 

[mansr]

Did you mean ATA/ATAPI when you said SATA?

SATA is an extension of the ATA8 spec. Does either of them allow indefinite completion times?

 

[Berwhale]

Does either of them allow indefinite completion times?

OK, I will Google it for you.

The answer is no. Not when NCQ is in the mix...

"Native Command Queuing (NCQ) is an extension of the Serial ATA protocol allowing hard disk drives to internally optimize the order in which received read and write commands are executed."

"...There is no mechanism in NCQ for the host to specify any sort of deadlines for an I/O"

From: https://en.wikipedia.org/wiki/Native_Command_Queuing

Do I win £5?

 

[mansr]

OK, I will Google it for you.

The answer is no. Not when NCQ is in the mix...

"Native Command Queuing (NCQ) is an extension of the Serial ATA protocol allowing hard disk drives to internally optimize the order in which received read and write commands are executed."

"...There is no mechanism in NCQ for the host to specify any sort of deadlines for an I/O"

From: https://en.wikipedia.org/wiki/Native_Command_Queuing

Do I win £5?

That's not the same as allowing commands to take a week to finish. Anyhow, I dug out the (S)ATA specs, and other than a special isochronous mode (probably what those video surveillance drives use), timeout conditions are mentioned as a possibility, yet bizarrely there doesn't seem to be a clear definition of when the host should declare one to have occurred, with or without NCQ. I guess it's all working by dumb luck.

 

[Ron Texas]

Music streaming does not tax the speed of a spinning drive. The main consideration is noise. Portable external drives are quiet and cheep. I use them.

 

[mansr]

https://arstechnica.com/gadgets/202...smr-disks-arent-great-but-theyre-not-garbage/

 

[somebodyelse]

https://arstechnica.com/gadgets/202...smr-disks-arent-great-but-theyre-not-garbage/

You beat me to it
The blog post from iXsystems has some further information - they're a NAS manufacturer using ZFS and they're working with WD to identify and resolve compatibility issues. There are preliminary indications that the fault condition causing the drives to be kicked out of the array may be limited to a specific firmware version and possibly drive size, which would explain why some are seeing outright failures while others report that it works but is relatively slow.
Edit: https://www.servethehome.com/wd-red-smr-vs-cmr-tested-avoid-red-smr/

 

[bravomail]

Plus silence for the SSD, if it matters to you.

SSD will have speed, silence, low power consumption and no mechanical moving parts. And since you will be using it as read-only - it should last forever. I'd take 2TB - for Windows, movies and games too. $200 - is not much.

 

[mansr]

SSD will have speed, silence, low power consumption and no mechanical moving parts. And since you will be using it as read-only - it should last forever.

There is no such thing as a read-only SSD. Once in a while, a few bits will be corrupted. When detected, either through normal read or a background scan, this corruption is corrected and the data rewritten. Modern multi-level NAND flash also exhibits a phenomenon called read disturb whereby simply reading a block of data can cause corruption. Again, this is automatically detected and corrected. Of course, these maintenance writes result in less wear than if the device is additionally being actively written to. Leaving an SSD unpowered for a long time (a year or more) is also a bad idea since accumulated corruption may be too much for the error correction to handle.

 

[mansr]

Elsewhere on the internet: https://www.computerbase.de/2020-06/nas-festplatten-synology-wd-red-smr-inkompatibel/

 

[somebodyelse]

A couple of articles with updates on the WD Red issue. The class action isn't after money - the remedy they're after is an injunction preventing WD from falsely advertising SMR drives as suitable for NAS use. There's also a roundup of NAS manufacturers putting the drives on their 'incompatible' lists, issues in non-ZFS use, a bit of who-knew-what-when from blog posts and NAS forums going back to last year, and WD's ongoing silence on the matter.

https://arstechnica.com/information...al-wants-to-end-any-use-of-smr-in-nas-drives/
https://www.servethehome.com/wd-red-dm-smr-update-3-vendors-bail-and-wd-knew-of-zfs-issues/