‘Geen verschil’ tussen goedkope en dure harde schijven

[ispam.nl]

Er is geen verschil tussen harde schijven die voor consumentensystemen worden gebruikt en professionele opslagsystemen. Dat stelt BackBlaze dat zelf 25.000 harde schijven gebruikt voor de online backup-service die het bedrijf aanbiedt. Het overgrote deel van de schijven zijn oorspronkelijk bedoeld voor de consumentenmarkt. Of de Nederlandse datacenters en hosters nu massaal kunnen besparen op […]

Gerelateerde artikelen:
Cloudstorageprovider: Na 3 jaar fors meer problemen met harde schijven
‘Harde schijven kunnen veel efficiënter werken met hitte’
Verkoop servers flink gedaald in vierde kwartaal 2011

Source: ispam.nl

[The-BosS]

Wat een vergelijking is dit, nergens op gebaseerd als je het mij vraagt. Kun je net zo stellen dat een lada ook minder mankementen heeft dan een dure sportwagen.

Op basis daarvan heeft BackBlaze een vergelijking gemaakt. Opmerkelijke uitslag: de goedkopere schijven doen het eigenlijk veel beter en hebben een jaarlijkse uitval van 4,2 procent. De professionele systemen doen 4,6 procent.

Dat is dus een verwaarloosbaar verschil, vooral omdat er blijkbaar niet eens een onderscheid gemaakt wordt tussen type en merken en ze het in de volgende quote alweer tegenspreken.

Na drie jaar ziet BackBlaze de uitval bij consumentenschijven wel toenemen. Dan zijn de professionele systemen natuurlijk wel in het voordeel omdat deze schijven een langere garantie kennen.

Ja maar wat is het nu, ofwel is er geen verschil ofwel is er wel degelijk een verschil.

[Domenico]

Misschien dat de originele blog of het artikel (1 op 1 kopie?) jullie wat meer info verschaffen.

http://lifehacker.com/why-enterprise...ost-1476333889

Op het blog gaan ze net iets dieper op de materie in.
http://blog.backblaze.com/2013/12/04...e-reliability/

[t.bloo]

Oud nieuws, is al lang een draadje over

http://www.webhostingtalk.nl/hosting...-schijven.html

[Domenico]

Oud nieuws, is al lang een draadje over

http://www.webhostingtalk.nl/hosting...-schijven.html

Klopt maar dit was een vervolgpost van Backblaze.

Last month I dug into drive failure rates based on the 25,000+ consumer drives we have and found that consumer drives actually performed quite well. Over 100,000 people read that blog post and one of the most common questions asked was:

“Ok, so the consumer drives don’t fail that often. But aren’t enterprise drives so much more reliable that they would be worth the extra cost?”

Well, I decided to try to find out.

[Mikey]

Vind het artikel op ispam mager, maar het bron artikel is ook mager. Er wordt gesproken over enterprise grade, dell verkoopt ook rebranded seagate schijven uit dezelfde sata lijn voor consumenten als enterprise. Dus in hoeverre is dit een eerlijke vergelijking ? In het bron artikel wordt er namelijk niet over gesproken, of het sas of sata is. Dan wordt er een conclusie getrokken door een defecte schijf uit de EMC.

Bij het bron artikel is er een comment die een goed punt heeft, enterprise sas bied je nog tal van meer voordelen ten opzichte van een consumer grade disk.

[Domenico]

Ik zal de comments hier even posten voor de luie mensen.



16 comments

Alan Brown · Follow
FWIW, my emperical experience with enterprise drives (and expensive scsi drives) over the last 30 years mirrors your more analytical study - in fact, if anything the expensive drives had a significantly higher failure rate than desktop hard drives (by a factor if at least 10!)
Reply · 1 · Like · Follow Post · 17 hours ago

Chris M Evans · Director at Langton Blue Ltd
Guys, what EMC array are you using? You should be aware that enterprise-class arrays will proactively spare a potentially failing disk even if it isn't failed. This makes RAID rebuild much quicker and safer as a RAID rebuild can hit an unrecoverable read error and so fail to rebuild data. Were your failures true failures or soft failures?
Reply · 1 · Like · Follow Post · December 4 at 11:52pm

Greg Knieriemen · Westlake, Ohio
Chris is absolutely right - there is a huge distinction between soft and hard failures - and completely invalidates your findings without the details.
Reply · Like · Edited · Yesterday at 3:46am

Alan Brown · Follow
Regarding your comments about vibration.... This begs the question: Is it better to rigidly hold a drive in a bay (and possibly couple vibration from adjacent ones) or is it better to put it in a mount with some damped compliancy in it so that coupling is eliminated? How can you then chose 60-80 drives in a large drive drawer and have everything work? Should you use backplanes with rigidly fixed connectors or should the connectros themselves be able to move with the drive, etc etc.

Is backblaze in a position to measure latencies, etc as in the classic "Don't shout at your disk drives" youtube video? (said video really does drive home the importance of vibration isolation.

I'm still watching, as I'd like to see your designs become sucessful enough to deploy in the datacentres I work in (I herd about 500TB of space science data and that's constantly increasing)
Reply · Like · Follow Post · Edited · 17 hours ago

Gleb Budman · Follow · Top Commenter · Backblaze
The vibration question is an interesting one. Certainly we've seen that reducing vibration helps. We initially reduced vibration by putting "anti-vibration sleeves" around each drives (basically high-end rubber bands). In the v3.0 of the Storage Pod, we both increased the pressure on the drives (adding rigidity), but also added dampening. That definitely seemed to help.

As for using the Storage Pods for your purposes, we store 75 petabytes on them, and hundreds of other companies/orgs use them as well, so they're ready to store your data. However, I would make sure the data is kept redundantly no matter what system you store it on.
Reply · Like · 11 hours ago

Alan Brown · Follow
The single largest probloem with consumer drive is the incredibly long "hang" times you get out of them when they encounter a bad sector. This used to be tunable (scterc) but the HDD makers realised there was money to be made and locked the parameter, forcing clients to use the newly introduced next model up ("NAS drives") - which is only different from consumer drivers in that the timeout paramater is tunable - This means consumer drives are difficult to use in RAID setups for nothing other than marketing reasons. (I used to use consumer drives all the time and have been forced to go upmarket simply so that RAID arrays don't go offline for 5 minutes at a time if there's a minor error.)
Reply · Like · Follow Post · 17 hours ago

Gleb Budman · Follow · Top Commenter · Backblaze
Alan, the drive companies have claimed that the enterprise drives have more to them than just the "hang" times - but on the reliability metric, if they are different, it doesn't seem to help.

As for the "hang" times, yes, that may impact performance, but from a storage reliability/density/cost perspective, the consumer drives are clearly a win - even in RAID arrays.
Reply · Like · 12 hours ago

Alex Narayan · Database Change Administrator at CorVel Corporation
Very interesting article, Brian. Would be even more interesting if you let us know what makers/models of consumer drives you have been using.

Also, you make no comments on "RAID/NAS" drives (eg, Western Digital's RED line, or Seagates NAS line). Are they somehow included in the study? If yes, are they included as "Enterprise" or as "Consumer" drives?

As most (all?) consumer drives nowadays have [no] way to set TLER/scterc, IMO it makes them completely unsuitable for array use. That's what makes RAID/NAS drives (which have TLER/scterc) so relevant for me...
Reply · Like · Follow Post · 19 hours ago

Gleb Budman · Follow · Top Commenter · Backblaze
Alex - stay tuned. We're digging in to see if we have enough data detail to make statistical determinations between drives. (On the enterprise drive set, my guess is "no", but on the consumer drives we should be able to.)

As for whether consumer drives are unsuitable for array use...that is what the drive companies claimed. However, 100% of our 25,000 consumer drives are in RAID arrays.
Reply · Like · 12 hours ago

Shawn Molnar · Vernon, British Columbia
That sample size is hardly a fair comparison, but I do see where you are coming from a reliability view. As well, It all depends on how you're using those enterprise drives. If they're dual port SAS and sitting in an enterprise array with redundant storage processors and connectivity all the way through to the drives, you get benefits such as online controller code upgrades without having to bring the array down and active-active failover. If you swapped those enterprise drives with consumer drives, you'd lose access to drives if you brought down the active controller. With enterprise drives and enterprise hardware, the other controller can pick up the path and not even drop a write op. Of course you do pay big $$$ to get these features, which is why they are enterprise solutions
Reply · Like · Follow Post · December 4 at 7:37pm

Gleb Budman · Follow · Top Commenter · Backblaze
Shawn you make a great point. If you need other features of the enterprise drives, you may wish to pay for them. However, from a pure reliability perspective, they aren't worth it. (FYI - we do use the enterprise drives in a enterprise Dell & EMC systems.)
Reply · Like · December 4 at 8:00pm
a2z7_80920 (signed in using yahoo)
What you are paying for is the intelligent (i.e. expensive) software around that disk. I bet that everyone in this forum understands that an "enterprise" drive installed in a home environment will have the same failure rate as any other drive. Afterall, that is what RAID stands for. Enterprise means very high and/or very redundant I/O delivery. Comparing one component of an enterprise system just tells me the writer doesn't understand how to move fast database I/O traffic.
Reply · Like · Follow Post · 21 hours ago

Gleb Budman · Follow · Top Commenter · Backblaze
The enterprise servers are ALSO orders of magnitude more expensive than Backblaze Storage Pods, and we pay for those for a variety of reasons. But the drives themselves are more expensive and should only be if the drives themselves are better - which the data shows them not to be.

As for moving fast database I/O traffic - that's not a criteria we care about for the majority of our storage. The question for us is whether the reliability of the enterprise drives is higher.
Reply · Like · 12 hours ago
rperez414 (signed in using yahoo)
This study is like comparing a Ford Fiesta to a BMW 540i and saying that after 3 years of usage the BMW's are worse. Whereas, the real delta won't show up for another decade. Or cheap windows compared to expensive windows for a short period of time. What a worthless article. I want my 2 minutes back.
Reply · Like · Follow Post · 9 hours ago

Cameron Bayly · Marketing Manager at NIRIX
Interesting article. You have one small typo though in the second last paragraph. Should be "...the longer warranty is less than..." not "... the longer warranty is less that...".
Reply · Like · Follow Post · December 4 at 7:25pm

Brian Beach · Distinguished Engineer at Backblaze Online Backup
Thx. Fixed.
Reply · Like · December 4 at 7:40pm

[dicktump]

Het valt mij hier wel op dat de Seagate consumentendrives een fors hogere uitval hebben dan bijv. de WD RE4 drives. Maar dat kan misschien ook aan het specifieke type van Seagate liggen. De WD RE4 disks zitten qua uitval rond het niveau van de SAS disks die we gebruiken, tot nu toe. Dus zelf zie ik de bevindingen van BackBlaze hier niet terug, als in dat enterprise disks wel iets betrouwbaarder lijken te zijn.

Uiteraard zijn we door o.a. RAID niet afhankelijk van de uitval van disks, maar goedkope disks die vaak uitvallen kunnen aardig duur worden als er telkens tijd in de vervanging en RMA procedure moet worden gestoken. Dus dan liever duurdere disks