Jump to content

SMR drive performance with docker services


drumstyx

Recommended Posts

8TB drives have just broken the price threshold to be cheaper per GB here in Canada (albeit only the externals, so they must be shucked), so while 4TB is $113 for an OEM drive (Seagate BarraCuda), I'm considering $180 for a Seagate Expansion 8TB (STEB8000100), which likely contains an archive/SMR drive.

 

Now, I know the performance for general NAS use has been proven alright, but of course tests show VERY slow performance on smaller files.

 

I run Plex, OpenVPN, Sonarr, Radarr, and a number of other docker images to support them, and when I look at read/write/s at any given time, there are a few KB moving around pretty much always. I figure this is just what happens when you run Plex and such, since they all run automated cleaning/processing.

 

How does the above scenario play with SMR? Would I get a noticeable slowdown? Part of me just wants to say screw it and get some 4TB drives -- I'd need to buy 2x8TB drives first anyway, just to fill my parity...

Link to comment

I don't think you're likely to find an Archive drive inside on of the current Seagate Expansion units. You're much more likely to find a BarraCuda Compute drive, which is also based on SMR technology but is somewhat inferior, having a particularly low annualised workload rate of only 55 TB per year, which you'll exceed by a considerable margin just by running a monthly parity check.

 

Slow performance on small files is not a feature specific to SMR. It affects all drives because there's a fixed file system overhead associated with writing a file. So if you write a single 100 GB file or a million 100 KB files you're writing the same amount of data but there's a million times the overhead with the latter and that accounts for the drop in performance.

 

I run a Plex docker. The docker image and the Plex database and metadata are stored on an SSD cache pool. That's what most people do and it works well. Only the large media files themselves are stored on the array. I don't run downloaders but if I did I would assemble the downloaded files on the cache pool and only move them to the array when they are complete.

Link to comment

Woah, that's actually a shockingly low workload rating...

My alternative to this drive was going to be the 4TB BarraCuda for $113, but looking up the specs it's just as bad.

Man, when did hard drives get so crap?

 

EDIT: Wait a sec, isn't that just write workload?

Link to comment
12 hours ago, drumstyx said:

Woah, that's actually a shockingly low workload rating...

 

Not for the purpose for which is was designed - to be left in its external case and used as an expansion/backup drive. You can get bargains by shucking but you have to choose wisely and this isn't one of them.

 

12 hours ago, drumstyx said:

My alternative to this drive was going to be the 4TB BarraCuda for $113, but looking up the specs it's just as bad.

 

Get the IronWolf instead. It's designed for the purpose you have in mind.

 

12 hours ago, drumstyx said:

Man, when did hard drives get so crap?

 

When people stopped wanting to pay for them.

 

12 hours ago, drumstyx said:

EDIT: Wait a sec, isn't that just write workload?

 

No. Total read/write workload. Why would writing wear the drive mechanism but reading not cause any wear at all?

Link to comment

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...