Seagate Ships World's First 8 TB Hard Drives

[newtekie1]

heard of RAID5 or RAID6? makes better use of the drives

Read my sig.

 

[MilkyWay]

Imagine trying to access 8tb of data on a HDD, i'd much rather raid together smaller drives. The only purpose i can see for single 8tb drives at the moment is data backup, reminds me of tape backups which where also slow as molasses.

 

[The Von Matrices]

Imagine trying to access 8tb of data on a HDD, i'd much rather raid together smaller drives. The only purpose i can see for single 8tb drives at the moment is data backup, reminds me of tape backups which where also slow as molasses.

Most likely these drives will use higher density platters, which means higher transfer speeds than existing drives, somewhat mitigating the issue.

That said, while the number of sectors per cylinder is increasing (which results in increased transfer rate), the number of cylinders per drive is also increasing as areal density increases. Since the drive can only read from one cylinder per platter, that means the time to empty a drive increases as the number of cylinders increases.

I fear that at some point writing or reading an entire drive will become a day-long affair. This would make most types of RAID worthless because the chances of the array failing due to a second drive failing during a rebuild would be very high. I remember having a 40GB drive back in 2000 that wrote at 30MB/s, which meant I could read the whole drive in about 22 minutes. Now the 4TB drives in 2014 transfer at ~180MB/s, which means that reading the whole drive takes about 370 minutes. By this logic we only need ~250TB drives before reading the entire drive takes longer than a whole day.

 

[newtekie1]

You don't have to write the entire drive to do a RAID rebuilt, just the part with data on it. Though with volumes this big, RAID6 is highly recommended.

 

[The Von Matrices]

You don't have to write the entire drive to do a RAID rebuilt, just the part with data on it.

Only if your drive is partitioned by the RAID controller to not use 100% of its available capacity, which seems like an unlikely scenario.

The RAID controller is dissociated from the file system. Without knowing which sectors are relevant data and which are deleted data, the RAID controller has to reconstruct every sector, deleted files and all.

 

[newtekie1]

Only if your drive is partitioned by the RAID controller to not use 100% of its available capacity, which seems like an unlikely scenario. The RAID controller is dissociated from the file system. Without knowing which sectors are relevant data and which are deleted data, the RAID controller has to reconstruct every sector, deleted files and all.

With RAID5/6, the parity is used to reconstruct the data that should be on the replacement drive. So only the part with actual data is written, not every sector. With RAID1, only the occupied sectors are copied, again not the entire drive.

Yes, that does mean that sectors that are occupied by deleted file data will be rebuilt/mirrored, since we know deleting a file doesn't actually delete the file from the drive just the pointer in the file table. But that is why it is a good idea to run a tool like the Wipe Free Space tool in CCleaner to wipe the free space on the volume. That way time isn't wasted rebuilding/mirroring data that has been deleted.

But, again, you don't write the entire drive during a rebuild, just the part with data on it, even if that data is garbage data.