0

I am looking to set up a RAID mass storage array, and have been reading about how large disk sizes are jeopardising rebuild because of the time required to execute a full rebuild and the chance of failure in the process. Research suggests the knee-point for RAID reliability is at about 1TB per drive for an 'acceptable' security of all data.

I am therefore looking to set up a system with several 1TB drives and am looking at the Seagate Barracuda ST1000DM003 drive. Looking at its spec sheet however, the 1TB drive is simply one third of a 3TB drive inside, i.e. one platter per TB, three in total for the 3TB, one platter for the 1TB. It seems like a 3TB drive is just three 1TB drives in the same space.

Given all that, is it possible to 'see' a 3TB drive as 3 1TB drives, one per platter, and having them fail per-platter rather than per disk. It seems this would either be easy, just choose the right cylinder/head/sector set for each platter, or impossible because the disk hardware abstracts these details away from the OS. Of course if the controller board fails, then that amounts to three disks going out which is bad, but in certain use cases doesn't make it impossible to use effectively.

Edit: I am discussing RAID 5 (or 6) with the parity.

2
  • 2
    What you've read about problems during rebuild are related to parity-based RAID levels like 5 and 6. RAID levels 0, 1 and 10 don't have this (potential/feared) problem. Use RAID 10 (and do regular backups) if you're worried about it. Keep in mind, redundant RAID levels are about uptime in case of hardware failure, not backup. Nov 18, 2014 at 17:13
  • 2
    A physical drive is partitioned into logical drives by allocating cylinders for best performance. An allocation by platter (actually surfaces and R/W heads) would be the worst-performance choice. Your idea is completely unreasonable because RAID expects independent and simultaneous operations among drives. Your "1 as 3" drive cannot perform 3 independent operations at once.
    – sawdust
    Nov 18, 2014 at 19:39

3 Answers 3

2

I believe you are oversimplifying how a hard drive functions and trying to apply a higher level of logic to solve a theoretical dilemma.

Hard drives are incredibly complex, and for decades the bits on a platter relationship to the bits you read and write was demolished. All modern drives do lots of work behind the scenes and deliver to you a virtual memory mapped linear structure. Sometimes divvied up into Cylinders and Heads, but that is entirely arbitrary and bears no resemblance to the actual drive structure.

Platters are not Drives. RAID is two or more DRIVES in a complex configuration designed to provide enhanced performance and/or reliability.

I have never heard of a 1 TB optimum for RAID. Check the source (and date). I wonder what the managers of petabyte arrays would say about the 1 TB limitation.

My experience is more along the lines of size your partition to fit your backup strategy. Some utilities are limited by the size of the backup medium, and today that limit is about 4 TB for USB. I have an 18 TB array that is giving me fits because I cannot back up the volume to a single drive, I am stuck doing a file by file.

1
  • 1 TB optimum is mainly quoted for RAID 5 arrays due to it's single parity drive. That means you will be running "without a safety net" the entire time it rebuilds (which could take several hours). The managers of those betabyte arrays would agree that you should not use RAID 5 and instead use something that has more redundancy than a single drive. Nov 18, 2014 at 23:37
1

As a direct answer to your question, No, it is not possible to view different HDD platters as different drives.

With that said, I agree with what @Techie007 and @timbo have said. I too have not heard of a 1TB optimal drive size (source?), and just as Techie007 said, rebuilding an array is only an issue if you're using a parity based RAID level. You don't mention what RAID level you're using.

To respond to your question about having the drives fail per platter instead of per disk, this is a risk you wouldn't want to take anyways. The same mechanics run all the platters, they experience the same amount of heat, vibration, etc. If one platter were to fail, it's extremely likely the others would fail as well. For example, if the motor goes out, all the platters are dead, same with the circuit board, arm, etc.

1

What I think you might be trying to research is the optimization of Multiple Head Hard Drives.

Check out this article on SuperUser:

Multiple independent read/write heads on the same hard drive platter?

However, all that being said, for this type of custom configuration, many controllers offer specific features and optimizations for specific uses, i.e., read only, write mostly, both read/write/rewrite optimizations.

For the most part, trust the Controller and Hard Drive manufacturers, and their drivers, or talk directly to the vendors for high end optimizations.

I have actually had custom controllers provided by Western Digital and Adaptec for SAN Solutions. Its kinda neat what they can come up with. Oh--and that costs, a lot.

1
  • So ultimately they are not willing to open up the HDD internals even an inch - the only option is to pay for them to do the specific optimisation that we want. I thought with HAMR, SMR and PMR, the older drives would be ripe targets for opening up at a lower level.
    – Milind R
    Oct 10, 2016 at 8:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .