Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
unknown
Nov 16, 2002
Ain't got no stinking title yet!


Just want a confirmation: SSDs start slowing down once you've written a shitload to them, right?

Backstory: Have an app that writes to a circular buffer at ~5MB/s 24x7 to a WD 1TB Blue 3d (and has for like a year and a bit) - only now starting to see (minor) spikes in latency on occasion that haven't really seen before when the write queue goes big. Smart results are perfectly fine - and it looks like the WD ncache buffer tech which uses SLC is almost perfect for this app. (Smart reports 5TB written to TLC, and 62TB to SLC, host writes of 56TB)

unknown fucked around with this message at 20:48 on Jan 29, 2019

Adbot
ADBOT LOVES YOU

unknown
Nov 16, 2002
Ain't got no stinking title yet!


Thanks, that makes sense - drive is like 80% empty, and there's no bad block/realloc errors at all in smart.

But with the huge cache hit ratio I was wondering if the SLC/ncache portion of ~15GB (described here: https://www.techarp.com/reviews/1tb-sandisk-ultra-3d-ssd-review/2/) was getting smashed apart. But further research puts SLC at 90-100k writes before an issue. (So like 125-150TB)

It's probably something else throwing poo poo into the write queue causing the delay - I am talking about a 1ms "spike" here.

unknown
Nov 16, 2002
Ain't got no stinking title yet!


Klyith posted:

AFAIK the SLC isn't a static area of the NAND, it gets moved around to work with the overall wear leveling. Newer drives don't even have a fixed size allocated to SLC mode. Writes in SLC mode are much less damaging to the cell, much like getting hit with a bat is nicer than getting shot with a bullet. The bat and bullet might have the same amount of kinetic energy, but the bullet is more concentrated.

Some SSDs *do* slow down over time with lots of data written. In many cases it isn't even the NAND itself wearing out, because doing a ATA Secure Erase will restore performance. OTOH some SSDs actually get faster!


So if you wanted to try to restore performance you could backup the drive and secure erase it. Alternately, I wonder if the constant writes are making the drive less efficient at its internal reorganization & garbage collection? If your application is able to be paused, you could stop it and run a trim periodically and see what that does.

Thanks for the data, quite interesting to read.

I wouldn't be surprised that there's some garbage collection type issues happening - the WD Blue 3D is literally a consumer level drive that was thrown in as a test, and this workload is really perfect enterprise type (solid writes - literally the os disk cache covers the reads I think) - probably should just spring for a samsung pro. Now that I think of it, I'm wondering if I can just mount a ramdisk for this app.

unknown
Nov 16, 2002
Ain't got no stinking title yet!


Klyith posted:

AFAIK the SLC isn't a static area of the NAND, it gets moved around to work with the overall wear leveling. Newer drives don't even have a fixed size allocated to SLC mode. Writes in SLC mode are much less damaging to the cell, much like getting hit with a bat is nicer than getting shot with a bullet. The bat and bullet might have the same amount of kinetic energy, but the bullet is more concentrated.

Some SSDs *do* slow down over time with lots of data written. In many cases it isn't even the NAND itself wearing out, because doing a ATA Secure Erase will restore performance. OTOH some SSDs actually get faster!


So if you wanted to try to restore performance you could backup the drive and secure erase it. Alternately, I wonder if the constant writes are making the drive less efficient at its internal reorganization & garbage collection? If your application is able to be paused, you could stop it and run a trim periodically and see what that does.

Just an update - looks like it wasn't the ssd at all, but an update was rolled into the app (written in java no less) that was screwing up more and more (probably some failed garbage collection) giving a false drive alert.

So, for people keeping track - a consumer level SSD (WD Blue 3d - 1TB) with a 24/7 writing stream to it for over a year is perfectly happy with the abuse. Smart reports are still saying 100% life left in it with zero errors on the drive.

unknown
Nov 16, 2002
Ain't got no stinking title yet!


https://www.theregister.co.uk/2019/11/25/hpe_ssd_32768/

https://support.hpe.com/hpsc/doc/public/display?docId=emr_na-a00092491en_us

quote:


SUPPORT COMMUNICATION - CUSTOMER BULLETIN
Document ID: a00092491en_us

Version: 1

Bulletin: HPE SAS Solid State Drives - Critical Firmware Upgrade Required for Certain HPE SAS Solid State Drive Models to Prevent Drive Failure at 32,768 Hours of Operation
NOTICE: The information in this document, including products and software versions, is current as of the Release Date. This document is subject to change without notice.
Release Date: 2019-11-19

Last Updated: 2019-11-22

DESCRIPTION
IMPORTANT: This HPD8 firmware is considered a critical fix and is required to address the issue detailed below. HPE strongly recommends immediate application of this critical fix. Neglecting to update to SSD Firmware Version HPD8 will result in drive failure and data loss at 32,768 hours of operation and require restoration of data from backup in non-fault tolerance, such as RAID 0 and in fault tolerance RAID mode if more drives fail than what is supported by the fault tolerance RAID mode logical drive. By disregarding this notification and not performing the recommended resolution, the customer accepts the risk of incurring future related errors.
HPE was notified by a Solid State Drive (SSD) manufacturer of a firmware defect affecting certain SAS SSD models (reference the table below) used in a number of HPE server and storage products (i.e., HPE ProLiant, Synergy, Apollo, JBOD D3xxx, D6xxx, D8xxx, MSA, StoreVirtual 4335 and StoreVirtual 3200 are affected. 3PAR, Nimble, Simplivity, XP and Primera are not affected.)

The issue affects SSDs with an HPE firmware version prior to HPD8 that results in SSD failure at 32,768 hours of operation (i.e., 3 years, 270 days 8 hours). After the SSD failure occurs, neither the SSD nor the data can be recovered. In addition, SSDs which were put into service at the same time will likely fail nearly simultaneously.

To determine total Power-on Hours via Smart Storage Administrator, refer to the link below:

unknown
Nov 16, 2002
Ain't got no stinking title yet!


Mental Hospitality posted:

I probably should have asked here before spending money...

Group opinion on the Western Digital Blue 1TB?

I have a whole bunch of them at work that are being used as cache drives for applications that are basically a 25/75 read/write ratio (yes) running 24x7. So basically the worst thing you can do to them. Not one bad one yet over the past year and smart reports are still giving me the 2 thumbs up.

unknown
Nov 16, 2002
Ain't got no stinking title yet!


Thought this might be interesting benchmark for people on a new system upgrade (Ryzen 3600+B550 motherboard) for me - 3 generations of drives: NVMe/WD SN850 (@pcix4 - max:7000/4100MBps), SSD/WD Blue (@sata6 - max: 560/530MBps), and HDD/Seagate Ironwolf (max:180MBps) .

Only registered members can see post attachments!

unknown fucked around with this message at 20:28 on Mar 22, 2021

unknown
Nov 16, 2002
Ain't got no stinking title yet!


I enabled bit locker on my sn850 recently and it was fun to watch it encrypt at 1 GByte/s (vs 125MByte/s for a sata ssd), but that's about it for performance improvements on day to day. There's maybe a couple of seconds shaved off here and there for loading stuff (email client and some games), but that's about it.

Adbot
ADBOT LOVES YOU

unknown
Nov 16, 2002
Ain't got no stinking title yet!


Just for people's curiosity about longevity - here's the smart output from a WD Blue 3D 1TB drive that's getting brutalized as a caching disk with a constant write of around 1-2MB/s to it for the past 33,000 hours (ie: almost 4 years - when they came out).

Yeah, don't worry about them wearing out.

Only registered members can see post attachments!

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply