4Kn or 512e server SAS drives?

[Kremlar]

Working on a new server configuration that will be running Windows Server 2012 R2, will be using Hyper-V to host several VMs. VMs running Windows Server 2012 R2 as well. Will be hosting various Windows applications, including Exchange, software that utilizes SQL, etc.

I notice that Seagate offers 2.5" 15K RPM drives that are 4Kn or 512e. In a server configuration where drives will run off a modern RAID controller does this matter? Is there any advantage to going with 4Kn drives instead of 512e?

I'm planning on using the Intel S2600CW2 motherboard with RS3DC080 RAID controller. I'm also planning on going with self encryption drives with the AXXRPFKDE2 encryption key.

For example, I'm looking at these Seagate drives:
ST600MX0102 (4K)
ST600MX0072 (512e)

Price is the same.

Any tips appreciated. Thanks!

 

[dave_the_nerd]

Shouldn't be any practical difference, as long as your RAID controller, OS, etc., support the larger sectors.

 

[Kremlar]

That's my understanding too. So, question is, why go with a 4Kn drive then? Any real reason to or is it safer to stick with 512e?

 

[CiPHER]

Drives with native 4K sectors cannot have alignment problems, and do not need >32-bit LBA support or GPT partitions to go beyond 2TiB addressing/partitions.

512e are physically 4K but emulate 512-byte sectors. They are compatible with all legacy software, but when misaligned will degrade in performance.

 

[Kremlar]

Thanks for the reply. When you say software, do you mean the OS?

In a RAID environment, running Hyper-V, does the software on a VM know or care if a drive is 512e or 4Kn?

If not it seems to make sense to go with 4Kn drives.

 

[CiPHER]

No I mean the software running on the OS, like several tools that simply have 512-byte sectorsize hardcoded in them because for a very long time this was the only sectorsize one needed to support.

BSD UNIX does know about the (native) sectorsize even if the disk says 512-bytes it will know it is 4096 bytes physical. But this is quite rare and most operating systems do not have such functionality. In a VM-solution, it would depend how the disk is presented. With vt-d and passing the entire controller this should be no different from running the guest VM without virtualization. But in most other cases the emulation layer would obfuscate the real sectorsize.

 

[zir_blazer]

Does they have to be 15000 RPM SAS HDs? I mean, if you want high end expensive enterprise HDs for performance instead of merely reliability, I'm sure that those HDs would get stomped by SSDs in every metric. Worse, high RPM SCSI/SAS HDs also had low capacity contrains, so they could get defeated by SSDs even on capacity.
I don't know if enterprise SSDs are in your price range or you could use multiple consumer SSDs with redundancy. But I don't think that high end SAS HDs are a sensible purchase.

 

[Kremlar]

Does they have to be 15000 RPM SAS HDs? I mean, if you want high end expensive enterprise HDs for performance instead of merely reliability, I'm sure that those HDs would get stomped by SSDs in every metric. Worse, high RPM SCSI/SAS HDs also had low capacity contrains, so they could get defeated by SSDs even on capacity.
I don't know if enterprise SSDs are in your price range or you could use multiple consumer SSDs with redundancy. But I don't think that high end SAS HDs are a sensible purchase.

Hi -

We need the reliability as well as the performance. Price for 600GB+ enterprise SSDs is very high. Also not mentioned in my original post, but we need a SED model preferably FIPS 140-2.

I'm not aware of any enterprise SSDs that have the capacity, price and encryption that we need.

 

[gvandehy]

Don't do it!!

Just implemented one with same configuration you are speaking of. Thought I did my research well, but what Microsoft says they support, doesn't mean it will actually work . . .

Installed Hyper-V host and 2 guests, all was well for a while. Then started getting random crashes in wpprecorder.sys, rebooted OK, then more frequent crashes [in the Host only], then boot failure.

There are many resources about this now that I know the errors. Essentially, 2012R2 [and Win 8.1 by association] cannot correctly decompress the *.sys files in windows\system32\drivers folder which leads to system crash.

The work around - so far - is to manually decompress the files and set an fsutil command to prevent automatic compression after windows updates.

Just got back from service call today on it. Looks stable now, but there are many compressed files that may still play a factor.

For reference, see MS KB 3027108, and a social.technet.microsoft.com thread https://social.technet.microsoft.co...thandled-wpprecordersys?forum=winserverhyperv

So far, the culprits have been cdrom.sys, and intelppm.sys.

My plan is to replace the 4k drives with 512b drives for the Host.

So far having the 4k drives for the guests hasn't triggered an issue

Hope this helps!

 

[Kremlar]

Wow, thanks for that link. That's reason enough to go with 512e!

 

[dave_the_nerd]

I know my advice is suspect because I work for a storage company, but why would you use NTFS compression on a production system? Cold storage, yeah, but a boot volume on a client-facing server? Seems suboptimal.

 

[XavierMace]

I know my advice is suspect because I work for a storage company, but why would you use NTFS compression on a production system? Cold storage, yeah, but a boot volume on a client-facing server? Seems suboptimal.

THIS.

 

[Kremlar]

I know my advice is suspect because I work for a storage company, but why would you use NTFS compression on a production system? Cold storage, yeah, but a boot volume on a client-facing server? Seems suboptimal.

Based on what I read, isn't Windows compressing those files automatically? I don't believe he specifically enabled compression on anything.

 

[gvandehy]

Based on what I read, isn't Windows compressing those files automatically? I don't believe he specifically enabled compression on anything.

You are correct. Compression was not selected. It is automatic / part of Windows Update process from what I understand.

The phrase that came to mind was the "Microsoft Scavanger" automatically searches for files to compress after windows updates. Thus the reason it wasn't problematic from the start.

Using the fsutil command disables that feature, so hopefully it doesn't re-compress those files [or any others]!

 

[gvandehy]

Wow, thanks for that link. That's reason enough to go with 512e!

I haven't tested this by any means, but as I mentioned, so far having the 4k for the guests hasn't triggered any issues. . . .

I have a RAID1 for the host OS and a RAID1 for the guests. Perhaps, and you would want to test this, you could run 512e for the host, and then 4Ks for the guests if you are looking for a little improvement??

 

[Kremlar]

I'm not convinced there is much of any improvement to be gained. As far as I can tell Windows Server 2012 R2 should be Advance Format aware so I should not have alignment issues with a 512e drive. Why Seagate is shipping 4Kn 600GB drives is beyond me. Only reason I can think of is, as a coworker said, they are using them as a field beta test!

Curious, you said you ended up with the host failing to boot. How did you resolve? Did you have to restore from backup?

 

[gvandehy]

I'm not convinced there is much of any improvement to be gained. As far as I can tell Windows Server 2012 R2 should be Advance Format away so I should not have alignment issues with a 512e drive. Why Seagate is shipping 4Kn 600GB drives is beyond me. Only reason I can think of is, as a coworker said, they are using them as a field beta test!

Curious, you said you ended up with the host failing to boot. How did you resolve? Did you have to restore from backup?

Once if stumbled across the link above, I used repair console to decompress cdrom.sys [not the wpprecorder.sys <= that was just a symptom]. Once I decompressed cdrom.sys I was able to normal boot and work thru the rest.

As a follow up, just heard from the end-user. Up and stable for 20 hours. This shouldn't be such a relief, but it is

Also, previously backups were failing due to VSS writer errors. Those are gone today as well and the backup was successful.

Not convinced we are out of the woods yet, and have the 512e drives on order in case I need to swap them out.

 

[Kremlar]

Good luck to you, and thanks again for the info.