HFS and drive sizes

#1 Sun, 6 Mar 2011 - 17:06 68kMLA

At what drive size does HFS begin to seriously impact storage and, to that extent, performance?

2GB? 4GB? 9GB?

#2 Sun, 6 Mar 2011 - 21:04 68kMLA

Storage wise, IMHO, 1GB. That's about the point you end up with minimum 32K block sizes if I remember correctly.

Though I have absolutely no clue what you mean by "seriously." 32K is pretty damn wasteful if you have a lot of files.

Performance wise, it's a complicated question because System 7 typically runs faster than 8.1 regardless of the file system used. Thus it's hard to say which is faster assuming the same hardware.

#3 Sun, 6 Mar 2011 - 21:34 68kMLA

Actually, strike that. Storage wise, it's whatever point you START seeing 32K blocks. I just read that 1GB should have 16K blocks under HFS. Some may even think 16K is too much, but I could probably live with it.

#4 Sun, 6 Mar 2011 - 22:56 68kMLA

2GB = 32K block sizes under HFS standard.

So yeah... that's about the point it starts getting offically stupid IMHO.

#5 Sun, 6 Mar 2011 - 23:41 68kMLA

2GB has been my rule of thumb, owing basically to the maximum size of drive recognized under early versions of System 7. But this has evidently been a mistake, as that was a theoretical rather than practical limit, and the practical limit is actually a good deal lower. I suppose on reflection that even into the System 7.6 era, drives were mostly under or around 1GB in size (thus 1GB drives or thereabouts shipped even in most of the pre-G3 beige PCI Macs), so this makes sense.

#6 Mon, 7 Mar 2011 - 04:28 68kMLA

Right. You definately start seeing diminishing returns when using large hard drives and/or volumes pre-HFS+. 7.5 can recognize 4GB volumes, but with a block size of 64K (yuck!).

#7 Mon, 7 Mar 2011 - 22:04 68kMLA

Isn't it the case that big blocks (32KB +) on a fast disk (or one with a big cache) work well with applications that use large files? The effects of fragmentation are reduced as long as you don't fill up the disk?

#8 Tue, 8 Mar 2011 - 22:28 68kMLA

I kinda doubt it. Even if it did, that's what defragging is for.

Naturally, a faster HD with a bigger cache is a good thing, regardless. I gather that putting "Macintosh," "SCSI," and "high performance" in the same sentence is something of a joke though. The Apple implementation of SCSI was robust (as SCSI usually is), but never particularly speedy. Especially with old 50-pin drives.

It'll still bottleneck at the controller.

Of course, HFS+ is impressively fragmentation resistant as it is. Even with its characteristic small block sizes. Which is yet another reason to use it if you can.