On 09/29/2017 06:42 PM, Peter Corlett via cctalk wrote: > On Fri, Sep 29, 2017 at 01:08:24PM -0700, Fred Cisin via cctalk wrote: >>> On 09/29/2017 11:20 AM, Jon Elson via cctalk wrote: > [...] >>> Older BIOS firmware provided no means for the user to define the geometry of >>> a connected drive - just a list of predefined types, and those often maxed >>> out at far less than any 512MB limit. There were various software solutions >>> to get around it, though. >>> Of course operating systems had various limits on the maximum size of a >>> partition on top of that - e.g. I think it was 32MB in earlier versions of >>> MS-DOS. >> What are the current drive size limits? > The latest relevant standards seem to be ATA-6 for ATA, and SBC-3 for SCSI, > which have 48- and 64-bit logical block addresses (LBAs) respectively. > > ATA LBAs are *always* 512 bytes no matter what the physical sector size is, so > this sets a hard limit of approximately 144EB. This limit also implies that > sectors must be a multiple of 512 bytes (e.g. modern 4kiB-sectored disks) and > unusual sector sizes are not supported. > > SBC-3 does not specify the size of a block, but it notes that most disks > support 512 bytes with some also supporting 520 or 4096. Assuming 512-byte > logical blocks, the limit is 9.44ZB. > > Since the hard disk equivalent of Moore's Law seems to be running out of steam > at mere tens of terabytes, we probably won't need to raise either limit for a > while yet :)
What not mentioned is that LBA addressed drives have cache. With that sector size is meaningless as the track (most likely) or several are read into the cache. There is LRU applied to keep the media up to date and also insure the cache is not stale. With cache it makes doing 512byte blocks a trivial issue. Now we have SSDs... Whole new game same old protocols. Allison
