@flomer I would say this is a post-sales matter so I would go that route. You bought something and it is not performing to your expectations. Is it a hardware problem, software configuration or did you buy the wrong thing? It doesn't matter - you paid big money for it so get the vendor to sort it out.
@DustinB3403 said in 16TB spinning rust is here:
@JaredBusch said in 16TB spinning rust is here:
Because 2% more is a lot.
Especially at the scale in which they purchase drives.
Most of the desktops drives you see in their stats are small drives. They are not buying small drives anymore. Only 8 TB and up. So mostly enterprise drives.
With their volume they get a good price anyway so who knows if they pay 2% more or not. Also if you get a longer warranty that might offset the price difference a little bit.
This is how much data they have stored on each type of drive:
@Obsolesce said in 16TB spinning rust is here:
@Pete-S said in 16TB spinning rust is here:
I think you can divide 3.5" drives into three major categories:
- Desktop drives - Consumer 8/24 usage, 2y warranty, SATA - WD Desktop, Seagate Desktop etc
- NAS drives - "Pro/semi-pro" 24/7 usage, 3y warranty, SATA - WD Red, Seagate Iron Wolf etc
- Enterprise drives - Heavy 24/7 usage, 5y warranty, SATA or SAS - WD Ultrastar, Seagate Exos etc
I think large desktop drives has become a niche market since that is not what people buy. And if you look at 10TB or more, the NAS drives and the enterprise drives cost almost the same but enterprise drives always have 5 year warranty so...
Unless trends will change I think you'll see a lot more enterprise drives used by Backblaze. I think in general they will always pick the lowest cost per TB drives they can find in volumes they can buy.
Assuming everything keeps moving to the cloud, there are going to be huge volumes of hyperscale high capacity drives sold and less of everything else. Considering energy cost and density, hyperscale companies are always looking for the highest capacity drives.
WD Blue 3/year warranty
WD Black 5/year warrantyBoth are non-Enterprise drives, and labeled as Desktop drives.
The fact that Backblaze used 34,737 Seagate st4000dm000 desktop consumer level drives that had a failure rate of 2.13% is by no means surprising. Most of the drives they use are not enterprise drives.
They must do it like that because they must think it's cheaper to deal with failed hard drives for the bulk of that tier of data.
However, looking at the HGST Enterprise grade drives they used almost 10,000 of (hms5c4040ble640)... those had a significantly lower failure rate.
Perhaps it's cheaper to go with Desktop drives for certain tiers of data, dealing with the failure rate, and for other data tiers, they choose Enterprise drives with much lower failure rates.
WD Blue for HDDs only has 2y warranty and is not made in large capacities. https://documents.westerndigital.com/content/dam/doc-library/en_us/assets/public/western-digital/product/internal-drives/wd-blue-hdd/data-sheet-wd-blue-pc-hard-drives-2879-771436.pdf
Regardless of that Backblaze have in the past said they pick the cheapest drives because the lower drive replacement cost doesn't offset the higher cost for better drives. Maybe they do pick different drives for different uses but I can't see why they should really. They have enough redundancy regardless.
But as I said, just looking at the trend last 12 months or so, I think you will see more enterprise drives in their lineup because prices are basically the same. Looking at our own prices I see that we pay only 2% more for the WD Ultrastar 10TB compared to the WD Red 10TB.
If we compare the E5-2667 v4 and the 6146 v1, the technological difference is nowhere near as large as between the E5-2643 V1 and E5-2667 v4.
The Gold 6146 was introduced mid 2017 while the E5-2667 v4 was mid 2016. Both use 14nm. Both use DDR4. The 6146 has 12 cores over the E5-2667v4 8 cores but it doesn't have the TDP to match that so it can probably not run as many cores in turbo at the same time as the E5-2667v4.
So all said and done, I'm not sure the 6146 actually is faster per core when all cores are under very high load. When you run a single core task 6146 should be faster but then the CPU has the ability to go into the highest turbo frequency on that core.
There is also the whole meltdown/spectre problems that have had a huge negative effect on performance. So if the new systems have patches for this and the old one don't, it could put the new systems at a big disadvantage.
Small things like BIOS settings can make a big difference when you are tuning for maximum performance. If you look at HPE and Dell for instance they have recommendations on BIOS settings for HPC applications that go against common IT practice but are what is needed for maximum performance.
Have you been in contact with Supermicro and their HPC guys? I think you need experts to tune your cluster to maximum performance.
I think you can divide 3.5" drives into three major categories:
I think large desktop drives has become a niche market since that is not what people buy. And if you look at 10TB or more, the NAS drives and the enterprise drives cost almost the same but enterprise drives always have 5 year warranty so...
Unless trends will change I think you'll see a lot more enterprise drives used by Backblaze. I think in general they will always pick the lowest cost per TB drives they can find in volumes they can buy.
Assuming everything keeps moving to the cloud, there are going to be huge volumes of hyperscale high capacity drives sold and less of everything else. Considering energy cost and density, hyperscale companies are always looking for the highest capacity drives.
@Dashrender said in 16TB spinning rust is here:
@Pete-S said in 16TB spinning rust is here:
@Pete-S 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.
Most companies buy drives from their vendor so they have no control over what drive they are actually buying. They are buying an HPE or Dell drive - manufactured by Seagate, WD, Toshiba etc. Just like they are buying 256GB of HPE RAM - manufactured by Micron, Samsung, Hynix etc.
And there is no data that suggests that Seagate drives fails more than others.
Backblaze have some data that suggests that some Seagate models are more failure prone than other drives on average but they also have data that suggests that other Seagate models have less failure than the average drive.
Don't they also use primarily consumer grade drives?
No, it depends. They use more enterprise drives as they have gone to higher capacity drives so they have both.
In the latest stats, the most reliable drive from all manufacturers that they have in larger numbers (>1000 drives) is the Seagate ST10000NM0086 at 0.33% AFR. That's a Seagate Exos X10 enterprise drive - 2.5M hours MTBF and 5 year warranty.
Backblaze also says that drive reliability in general has gone up and 2018 has been the year with the lowest failure rate since they started.
@scottalanmiller said in Another VPS Kid on the block upcloud.com:
@Pete-S said in Another VPS Kid on the block upcloud.com:
@scottalanmiller said in Another VPS Kid on the block upcloud.com:
@Emad-R said in Another VPS Kid on the block upcloud.com:
Plain KVM is so mature and everywhere I am starting to think we should all huddle up and create our own VPS service. I think between all ML we have around 500 VMs or so
So like.... 1-2 servers
It would be interesting to know the vCPU to core count ratio on these cloud providers. I'm guessing 10-15 vCPUs on each real core.
So a 16 core server can have 160-240 VMs.
So 1-2 servers sounds about right
And you can often get way, way bigger core counts. Easy to get at least 64 cores on a single server.
Yes, a server with dual AMD Epyc 7551 for instance has 32 cores per CPU so 64 real cores total.
@scottalanmiller said in Another VPS Kid on the block upcloud.com:
@Emad-R said in Another VPS Kid on the block upcloud.com:
Plain KVM is so mature and everywhere I am starting to think we should all huddle up and create our own VPS service. I think between all ML we have around 500 VMs or so
So like.... 1-2 servers
It would be interesting to know the vCPU to core count ratio on these cloud providers. I'm guessing 10-15 vCPUs on each real core.
So a 16 core server can have 160-240 VMs.
So 1-2 servers sounds about right 
@scottalanmiller said in Need video recording software for Fedora:
@Pete-S said in Need video recording software for Fedora:
Video is a lot of work. If you want the end result looking remotely professional, don't record on a webcam.
I record via a Lumix Z300. Love it. Doing a ton with it in Europe right now.
I'm not surprised. That's a Panasonic and they make good video gear.
The problem with webcams are that they are made for a specific purpose and because of that they have an ultra wide lens. At webcam distance it will cause perspective distortion of the face.
A professional would never place a camera as close as a webcam and would instead shoot with a narrower focal length further away to get the same framing. The end result will look much better - even if a camera with webcam image quality was used.
This is what perspective distortion looks like:
Video is a lot of work. If you want the end result looking remotely professional, don't record on a webcam. You need something better that can do manual exposure and a custom white balance. And you need some lighting as well but most important is a really good sound recording.