Jump to content

  • Log in with Twitter Log in with Windows Live Log In with Google      Sign In   
  • Create Account

Photo

SMART used as a predictor of failure?

BackBlaze Smart HDD Failure

  • Please log in to reply
13 replies to this topic

#1 Spider99

Spider99

    Advanced Member

  • Members
  • PipPipPip
  • 443 posts
  • LocationWarrington, UK

Posted 09 October 2016 - 08:14 AM

Interesting article from backBlaze on SMART info

 

Not definitive but interesting

 

Wish they would learn how to present stats and % properly as it makes it difficult to figure out what the numbers mean!

 

https://www.backblaz...drive-failures/


  • Ginoliggime and LavetaJef like this

Tim

---------

DrivePool & Scanner on 2012R2 Essentials Server Gigabyte Z87x OC Force, i7-4790S, 32 GB, LSI 9211, HP SAS Expander, 27 Storage disks, 6 SSD Cache, ~115 TB Storage


#2 Christopher (Drashna)

Christopher (Drashna)

    Customer and Technical Support

  • Administrators
  • 8,208 posts
  • LocationSan Diego, CA, USA

Posted 09 October 2016 - 06:46 PM

So far, I think this is the only post that BackBlaze has released about drive failures that won't generate controversy or anger. 

 

This pretty much follows my personal experience, as well as manufacturer info.  So really .... not surprising. 


Christopher Courtney

aka "Drashna"

Microsoft MVP for Windows Home Server 2009-2012

Lead Moderator for We Got Served

Moderator for Home Server Show

 

This is my server

 

Lots of "Other" data on your pool? Read about what it is here.


#3 Spider99

Spider99

    Advanced Member

  • Members
  • PipPipPip
  • 443 posts
  • LocationWarrington, UK

Posted 09 October 2016 - 08:36 PM

ha yes they do get people going - think its part Marketing and part tweaking the established player

 

especially "you dont need nas/enterprise" drives for 24/7 usage - when most of their drives do very little everyday in comparison to a company files server etc


  • Christopher (Drashna) likes this

Tim

---------

DrivePool & Scanner on 2012R2 Essentials Server Gigabyte Z87x OC Force, i7-4790S, 32 GB, LSI 9211, HP SAS Expander, 27 Storage disks, 6 SSD Cache, ~115 TB Storage


#4 Christopher (Drashna)

Christopher (Drashna)

    Customer and Technical Support

  • Administrators
  • 8,208 posts
  • LocationSan Diego, CA, USA

Posted 11 October 2016 - 05:57 PM

Yup, exactly.  

 

My biggest issue is sample size. And then people quote them like their data is anything but horribly flawed. 


Christopher Courtney

aka "Drashna"

Microsoft MVP for Windows Home Server 2009-2012

Lead Moderator for We Got Served

Moderator for Home Server Show

 

This is my server

 

Lots of "Other" data on your pool? Read about what it is here.


#5 Spider99

Spider99

    Advanced Member

  • Members
  • PipPipPip
  • 443 posts
  • LocationWarrington, UK

Posted 11 October 2016 - 09:28 PM

Ha yes well if others woke up and stopped moaning and provided there own data - with proper analysis that the average non stats geek can understand people might get a better understanding of

 

a) data use - most peopl cant analyse data for toffee - although to here them talk -  :blink:

B) choose the right HDD/SDD for your situation


  • Christopher (Drashna) likes this

Tim

---------

DrivePool & Scanner on 2012R2 Essentials Server Gigabyte Z87x OC Force, i7-4790S, 32 GB, LSI 9211, HP SAS Expander, 27 Storage disks, 6 SSD Cache, ~115 TB Storage


#6 Christopher (Drashna)

Christopher (Drashna)

    Customer and Technical Support

  • Administrators
  • 8,208 posts
  • LocationSan Diego, CA, USA

Posted 12 October 2016 - 02:13 AM

I like statistics, but ... yeah, BackBlaze's are horrible. :)

 

But what I've found over the years, is that aside from a bad line of drives every once in awhile, all companies are pretty much equal. 

 

Expect a bathtub curve for drive failure.  And buy drives for their purpose (in a storage server, get a NAS rated drive, SERIOUSLY). And don't shell/shuck drives. And most importantly (to me), check the warranty period. 


Christopher Courtney

aka "Drashna"

Microsoft MVP for Windows Home Server 2009-2012

Lead Moderator for We Got Served

Moderator for Home Server Show

 

This is my server

 

Lots of "Other" data on your pool? Read about what it is here.


#7 Spider99

Spider99

    Advanced Member

  • Members
  • PipPipPip
  • 443 posts
  • LocationWarrington, UK

Posted 12 October 2016 - 03:18 AM

yes a nice 5 year warranty if you can get it - means any problems you get a new (or refurb) drive :) noticed some of the newer Seagate are only coming with 3 yrs for enterprise drives - poor deal

 

yes rule of thumb i use is - if they last 6 months they usually last a good long time but after 3-4 years - you have either run out of space or some humongous drive is now available for silly money before they all ware out

 

i see a 60TB SSD for enterprise has appeared ($10k) - if they last you can say good by to mechanical drives in 3-5 years - so much easier to add another chip or two of memory than cram another platter into a 3.5 hdd case

 

then we will be moaning about memory chip life :)


  • Christopher (Drashna) likes this

Tim

---------

DrivePool & Scanner on 2012R2 Essentials Server Gigabyte Z87x OC Force, i7-4790S, 32 GB, LSI 9211, HP SAS Expander, 27 Storage disks, 6 SSD Cache, ~115 TB Storage


#8 Christopher (Drashna)

Christopher (Drashna)

    Customer and Technical Support

  • Administrators
  • 8,208 posts
  • LocationSan Diego, CA, USA

Posted 13 October 2016 - 06:48 PM

Yeah, the longer the warrant, the better. But yeah, I've noticed that a majority of drives are 3 year warranties, at most, anymore. :( 

 

Also, I've found that the warranty period is a pretty good indicator of the estimated drive livespan.  It should last roughly 

 

 

Well, it's funny.  Samsung thinks spinning drives are dead.  A bit biased, considering they sold off their HDD line. Almost sounds like they're trying to convince themselves. 

 

That said, while capacities are growing, the question is when will SSDs and HDDs hit the same price per TB? Eg, when will it be worth switching over? 


Christopher Courtney

aka "Drashna"

Microsoft MVP for Windows Home Server 2009-2012

Lead Moderator for We Got Served

Moderator for Home Server Show

 

This is my server

 

Lots of "Other" data on your pool? Read about what it is here.


#9 Spider99

Spider99

    Advanced Member

  • Members
  • PipPipPip
  • 443 posts
  • LocationWarrington, UK

Posted 13 October 2016 - 10:47 PM

The other thing that people are ignorant of or dont understand is a ssd has a finite life - for write - or write endurance or TBW or......

 

"cheap" bottom of the range ssd's - either dont say or are very low - while the larger the drive the better the write endurance is

 

people are using them more and more - on the qnap forums as the more modern and newest nas's have ssd slots and now m.2(sata) - it appears more are being used - not significant overall but an indication people are starting to see them as more storage rather than fast os drives.

 

As ssd at consumer level of ~1tb are now affordable(i have two 960gb and one 750gb) - i think it will be only a couple of years or so - driven by samsung looking at the sizes its looking like 250/tb or 5-6x the mechanical equivalent - add in the speed factor and its not that far apart


Tim

---------

DrivePool & Scanner on 2012R2 Essentials Server Gigabyte Z87x OC Force, i7-4790S, 32 GB, LSI 9211, HP SAS Expander, 27 Storage disks, 6 SSD Cache, ~115 TB Storage


#10 Christopher (Drashna)

Christopher (Drashna)

    Customer and Technical Support

  • Administrators
  • 8,208 posts
  • LocationSan Diego, CA, USA

Posted 14 October 2016 - 09:20 PM

Don't forget that the type of NAND cell makes a HUGE difference in performance and longevity too (SLC, MLC, TLC, etc). 

 

 And yeah, I know I use my 120GB SSDs as write cache, and am absolutely trashing the drives. :) 

 

 

 

Though, a good thing to keep in mind, "blazing fast" describes not only SSD performance, but also how they die. :)


Christopher Courtney

aka "Drashna"

Microsoft MVP for Windows Home Server 2009-2012

Lead Moderator for We Got Served

Moderator for Home Server Show

 

This is my server

 

Lots of "Other" data on your pool? Read about what it is here.


#11 Spider99

Spider99

    Advanced Member

  • Members
  • PipPipPip
  • 443 posts
  • LocationWarrington, UK

Posted 14 October 2016 - 10:44 PM

yes Nand and Nand3d may well change the game again

 

Yes i bought some Sk Hynix SL301 - 50 for 256GB - in ssd terms throw away - they perform well in the 500 MB/s plus range as virtually any SSD's do now

 

Ive not had a ssd die in proper usage - do they really go read only as is touted or just curl their toes up 

 

The one ocz thats causing me a problem is that it comes and goes on the sata port - sometimes its there the puff its gone - ocz known fault i believe

 

i need to test it with scanner - but i suspects its a drinks coaster

 

One other i had did start showing less than 100% life so i just binned it as it was four years old and was my old system disk and had had a hard life

 

I have a couple of Kingston HyperX 960gb with one i hammer all the time with bluray rips / blu ray backups and its not missed a beat - with 17TB reads and 14TB writes in four months or so of use

 

 

Have you played with nvme yet?


Tim

---------

DrivePool & Scanner on 2012R2 Essentials Server Gigabyte Z87x OC Force, i7-4790S, 32 GB, LSI 9211, HP SAS Expander, 27 Storage disks, 6 SSD Cache, ~115 TB Storage


#12 Christopher (Drashna)

Christopher (Drashna)

    Customer and Technical Support

  • Administrators
  • 8,208 posts
  • LocationSan Diego, CA, USA

Posted 15 October 2016 - 05:24 PM

I've only had SSDs fail horribly. 

 

As for the not showing up thing, I've seen that on non-OCZ drives as well.  Apparently certain haswell based chipsets have issues with SSDs....... 

 

I've never had one run out of spare cells... 

 

 

And no, I haven't had a chance to play with NVMe.  Only my desktop supports it (and required a firmware update to do so).  But I haven't had a pressing need.  My next system will probably use NVMe for the system and write cache drives, though. 


Christopher Courtney

aka "Drashna"

Microsoft MVP for Windows Home Server 2009-2012

Lead Moderator for We Got Served

Moderator for Home Server Show

 

This is my server

 

Lots of "Other" data on your pool? Read about what it is here.


#13 Spider99

Spider99

    Advanced Member

  • Members
  • PipPipPip
  • 443 posts
  • LocationWarrington, UK

Posted 16 October 2016 - 11:36 AM

Thanks for the haswell tip that might be the issue - as my new server is haswell based - although the os drive is also an ocz but an older model - probably an different controller -- hmm

 

You are in for a treat with nvme - have three 950 pro's - two in my Skull Canyon and one for my boot drive on my main pc

 

[attachment=728:M2 results clean install with samsung drivers.PNG]

 

those speeds are impressive and you see them in real world even when you have filled it up a bit

 

The main thing to watch for is the heat - they get very hot even with low usage - i had mine initially in a covered m.2 slot - and 55-60c at idle and would shoot up to 70+ under load - so i moved it to an add on card which i could give better cooling to - i also removes the sticker and added several small ram heat sinks to the "card" and now i only have 35c idle :)

 

Tip if it does over heat it does throttle the speed but what you also get is  

[attachment=729:Capture.JPG]

notice the last smart entry - lots of errors - these are when it over heats and not anything to worry about if you can improve you cooling

 

I got that particular drive as a cheapo from Amazon warehouse - and it came with 10k+ of those - and observed that when it over headed the number would go up - having now sorted the heat issue the number has remained static - i guess somebody returned it missing the overheat issue

 

This is the card i used (came with my m/b) - https://www.amazon.c...=I11ZSV1OSI2A83

No drivers needed just plug it in and your bios should see it

 

There are several youtube videos claiming that these disk slow down with thermal throttling - but hey are mostly equivalent to a Backblaze blog in that the warning should "no brain cells were used in the making of this this video" - on the Skull i can copy say a dir of large video files between the drives - initially they copy at max speed but when the cache gets exhausted the speed drops to only 1300MB/s ( :lol:  ) which is the true speed of these drives under sustained load. Yes if they do get very hot the throttling may happen but not what you would believe on first viewing.

 

The new 960's have moved things along some more - i believe they have improved the cooling as well - but i will stick with these for now and probably catch up @ the 990 stage :)

 


Tim

---------

DrivePool & Scanner on 2012R2 Essentials Server Gigabyte Z87x OC Force, i7-4790S, 32 GB, LSI 9211, HP SAS Expander, 27 Storage disks, 6 SSD Cache, ~115 TB Storage


#14 Christopher (Drashna)

Christopher (Drashna)

    Customer and Technical Support

  • Administrators
  • 8,208 posts
  • LocationSan Diego, CA, USA

Posted 17 October 2016 - 06:17 PM

Very nice. :)

 

I do have an M.2 slot on my board, and I am using it. But with an AHCI based drive, not NVMe. 

 

But It's a Z97 system, so a firmware update and it supports NVMe on the M.2 port (and the SATA 0-1 ports, IIRC).   And I've already flashed it. :) 

 

The problem is that it's a Mini-ITX system. So no spare PCI-e port, and it only supports 2242 M.2 cards.   It's a great board, but ...

There is a reason I have a seperate SSD for games. :)

 

 

That said, i'm not surprised about the heat thing. And I'm paranoid/obsessive about lowering the temperature. 


Christopher Courtney

aka "Drashna"

Microsoft MVP for Windows Home Server 2009-2012

Lead Moderator for We Got Served

Moderator for Home Server Show

 

This is my server

 

Lots of "Other" data on your pool? Read about what it is here.






Also tagged with one or more of these keywords: BackBlaze, Smart, HDD Failure

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users