16TB spinning rust is here
-
@scottalanmiller said in 16TB spinning rust is here:
@Obsolesce said in 16TB spinning rust is here:
2% of 100k HDDs has got to be over $200k. The time that goes into replacing them, performance loss, etc can't be overlooked either, which is added costs on top of the drive costs.
In theory, from old BB studies, consumer and enterprise drives fail at roughly the same rates. Speeds are the bigger differences, and sometimes capacities. Often hard to directly compare as different drive types tend to be made for the different audiences.
I haven't seen those. In practice, it seems like a different story according to their charts.
-
@Obsolesce said in 16TB spinning rust is here:
@scottalanmiller said in 16TB spinning rust is here:
@Obsolesce said in 16TB spinning rust is here:
2% of 100k HDDs has got to be over $200k. The time that goes into replacing them, performance loss, etc can't be overlooked either, which is added costs on top of the drive costs.
In theory, from old BB studies, consumer and enterprise drives fail at roughly the same rates. Speeds are the bigger differences, and sometimes capacities. Often hard to directly compare as different drive types tend to be made for the different audiences.
I haven't seen those. In practice, it seems like a different story according to their charts.
Older blog, but BB stated that consumer drives outperformed enterprise drives on failure rates...
https://www.backblaze.com/blog/enterprise-drive-reliability/
-
@scottalanmiller said in 16TB spinning rust is here:
@G-I-Jones said in 16TB spinning rust is here:
@scottalanmiller what's the general consensus here? Do we not buy Seagate because they fail more? Genuinely curious.
Generally avoid them, yes. Big usage shops like BB have had issues with them, and repair shops have advised that they see the failure rates on them in the wild being very high.
In 3.5" SAS we have plenty of 60-bay JBODs out there filled with Seagate NearLine drives. Our failure rates over the five year solution life across the board are virtually nil.
We had one drive fail in a Dell MD3060e but who the chicken knows what drive was installed in there in the first place.
We avoid Seagate SAS SSDs due to firmware problems that we've seen in the same settings.
We've been running with HGST now Western Digital drives and JBODs and have seen similar virtually nil fail rates.
As far as 2.5" 10K SAS goes, we have lots of Seagate and HGST out there with, again, very little in the way of fail rates.
Hard Drives have become quite reliable as a whole in the settings we deploy them in.
We've not deployed spinning rust in a workstation/desktop/laptop for years now. So, that negates the consumer side of things. We just don't.
-
-
@Obsolesce said in 16TB spinning rust is here:
@PhlipElder said in 16TB spinning rust is here:
JBODs
Totally fine is its some kind of RAIN. Actually i'm looking at building a CEPH cluster, Dell&RedHat recommend using single-drive RAID0s
-
@Kris_K said in 16TB spinning rust is here:
@Obsolesce said in 16TB spinning rust is here:
@PhlipElder said in 16TB spinning rust is here:
JBODs
Totally fine is its some kind of RAIN. Actually i'm looking at building a CEPH cluster, Dell&RedHat recommend using single-drive RAID0s
Our Storage Spaces Direct (S2D) hyper-converged clusters use IT Mode SAS HBAs for the SAS/SATA disk setup. NVMe are straight-through by default.
-
Many systems today can use JBOD, the system itself takes care of any reliability needs. Exchange added that back in 2013.
-
@Kris_K said in 16TB spinning rust is here:
@Obsolesce said in 16TB spinning rust is here:
@PhlipElder said in 16TB spinning rust is here:
JBODs
Totally fine is its some kind of RAIN. Actually i'm looking at building a CEPH cluster, Dell&RedHat recommend using single-drive RAID0s
No it was about the 'term', nevermind lol.
-
@PhlipElder said in 16TB spinning rust is here:
Hard Drives have become quite reliable as a whole in the settings we deploy them in.
Yes, even drives that fail "a lot" doesn't imply that you are likely to ever see one. If the best drives fail .000001% of the time, and the worst drives fail 300% as often, that's still only .000003% of the time.
-
@Kris_K said in 16TB spinning rust is here:
@Obsolesce said in 16TB spinning rust is here:
@PhlipElder said in 16TB spinning rust is here:
JBODs
Totally fine is its some kind of RAIN. Actually i'm looking at building a CEPH cluster, Dell&RedHat recommend using single-drive RAID0s
If you add RAIN, it's not a JBOD
Then it is a RAIN. -
@Dashrender said in 16TB spinning rust is here:
Many systems today can use JBOD, the system itself takes care of any reliability needs. Exchange added that back in 2013.
Everything uses JBOD at some level.
