4-TB and 3-TB Hard Drive Speeds

James Board

Distinguished
Sep 17, 2008
25
0
18,530
I want to purchase a new hard drive, either 3-TB or 4-TB, and I care most about the average read/write speed for the first 1-TB of the drive. I'll have a separate partition for that first 1-TB, and then configure a RAID across multiple hard drives in the same way.

Seagate 3-TB drives seem to be the fastest according to THG Charts, but there's lots of different Seagate 3-TB drives out there. How can I tell, when I'm at the retail store, which is fastest? Also, how can I tell how many platters the drive has? Presumably, given a constant overall capacity and rotation speed, less platters means larger capacity platters, and which means higher data transfer speed.

 
Solution
Popatim's advice is worth it's weight in gold.

Some controllers might let you create a RAID on a small section of the disk, then let you use the rest as JBOD, but I've never seen a situation where this was a good idea.

As far as the variety of drives available, there are a few main things to look at:
Cache size - the bigger the drive cache, the better if you're handling larger files, 32 or 64MB is fairly standard, though it could be a few GB if it's a hybrid drive.
The platter count is one that's harder to divine, but in general a newer generation drive will have fewer platters than an older gen drive of the same data capacity. If you can get detailed technical specs, it might be included in there. Fewer platters also means less wear...
Your assumption is correct about data density and greater speed. A fast drive speed(rpm) along with a high density platter will give you good data throughput.

How are you doing the raid? You can't use parts of a drive for a raid. It sounds like you want to use 1TB for something and then use the remainder for the raid. This won't work. Set up the raid first, and then create the partitions.
 

James Board

Distinguished
Sep 17, 2008
25
0
18,530
I was planning on having either 2 or 4 3-TB drives, and creating a 1-TB partition on each drive and implementing a RAID-0 across those 1-TB partitions. The remaining 2-TB of each hard drive would have regular 2-TB ext4 partitions (this is on a Linux machine) and each of those 2-TB partitions could be mounted independently.

That was the plan, though I have no experience with RAID so I don't know if it's possible. Are you saying it isn't possible?
 
It's not possible. When you create a raid, it uses the entire drive. You can't add parts(partitions) of a drive to a raid. It's all or nothing.

Once you create the raid, then you can partition it, but the partitions will be across the raid and not on a particular disk.

Also be aware that using raid 0 is fine, but make sure you create a good backup plan, and implement it. If 1 drive fails, you will lose all the data(from both drives).
 

James Board

Distinguished
Sep 17, 2008
25
0
18,530


Are you sure about this? I was told the exact opposite. If you are correct, then you saved me lots of $$! Would it make any difference if I'm creating this RAID on a Linux machine, not a Windows machine?
 

James Board

Distinguished
Sep 17, 2008
25
0
18,530

Take a look at this site: https://raid.wiki.kernel.org/index.php/RAID_setup. It says you can either use the entire disk, or merely a partition to configure the RAID. Why do you think the RAID has to use the entire drive? Where did you get your information?
 

popatim

Titan
Moderator
hardware raid typically uses the entire drive.
What you seem to have planned is software raid (which is perfectly acceptable)

Just be aware that if you create a raid drive D: and E: F: G: H: are your extended partitions that if you send a file to D: any operations done on E: F: G: or H: will slow down the work D: is attempting to do because its still using a physical disk that is doing work already. Having drive try to do double duty just makes both operations go at less than half speed.

For example, you begin writting a large sequential file to you 4 drive raid0 and its cruising at 400MB/s. then you want to also transfer a file from your laptop to one of the extended partitions, drive E:
Now drive E:'s physically part of the raid0 so when it begins to write, the raid0 goes into a wait state, the E: head has to reposition to write some of that data (being a single drive it wont see 400mb/s), stop and reposition the head back to the raid0 partition to write a bit more, stop and reposition back to E: to write a little more... over and over dropping your speed on both operations to about maybe 70 MB/s (if you're lucky)

If you're concerned about drive performance, get 4 1tb drives in raid 0 and four more single drives in whatever size you want for the remainder (if you really want 4 more drives LoL)
 

Saberus

Distinguished
Popatim's advice is worth it's weight in gold.

Some controllers might let you create a RAID on a small section of the disk, then let you use the rest as JBOD, but I've never seen a situation where this was a good idea.

As far as the variety of drives available, there are a few main things to look at:
Cache size - the bigger the drive cache, the better if you're handling larger files, 32 or 64MB is fairly standard, though it could be a few GB if it's a hybrid drive.
The platter count is one that's harder to divine, but in general a newer generation drive will have fewer platters than an older gen drive of the same data capacity. If you can get detailed technical specs, it might be included in there. Fewer platters also means less wear on the motor, and the drive is able to reach operational speed sooner.
More platters also means more heads, which depending on the design, could mean faster data transfer because it's reading from 8-12 heads instead of 3 or 4.

If you're really wanting speed and have the money to burn, SSDs can and do outperform HDDs. There's also SSD PCI-e cards that cost a small fortune, but behave like SSDs in a RAID setup, and they communicate through the PCI-e bus without eating up SATA ports. You should do some analysis and see if either solid-state solution would be worth it to you.
 
Solution