Quote:

For sure.  I'm with Nigel on this one.

Pleonasm wrote on Feb 2^nd, 2009 at 8:46am:

I think that most benchmarks would agree with you.  But, Nigel isn't talking about how HD benchmarks work, he's talking about the way programs you use every day interact with your hard drive.  Despite improved benchmarks, most people don't notice a worthwhile speed improvement in everyday tasks by moving data to faster parts of the disk.  

Although, on F u r u y a's behalf, it seems he is a little new to all of this, and the knowledge he is gaining from the experience may be worth the trouble, no matter if he sees a speed improvement or not.  

F u r u y a wrote on Feb 3^rd, 2009 at 5:22am:

It's an incredibly bad idea to partition these disks at all, and the location of the files on the disk is almost certain to be completely immaterial on a task like video re-encoding for files in consumer size ranges; for that performance is typically dominated by the activity of the video codecs, not I/O throughput (in professional situations to address the real difficulty with genuinely large raw HD video - in the 100Gb+ range - of moving around that data, things like FiberChannel-based shared storage get used).

The innermost - the slowest - zones of a 2008-vintage 1Tb drive should be delivering around 60Mb/s of sustained read throughput, a speed that would have been considered exceptional for the outermost zones of a drive from 4 years ago, and the 2009 models due out shortly from WD and Seagate should have even higher sustained transfer speeds across the board.

Since this is a consumer and not a professional situation the only task you'll every do in which the sustained throughput performance will matter is copying those files around, and by excessive over-partitioning in pursuit of tiny micro-optimizations the more of that you'll force yourself to do - so you end up spending all your time manually shuffling data instead, which will cost you more than any gain you think you might get from second-guessing drive performance. It's a penny-wise, pound-foolish approach.

F u r u y a wrote on Feb 3^rd, 2009 at 10:17pm:

My aim is not to convince; merely making sure that baseline accurate information is at least presented once by someone.

[ It may be on occasion that it is reasonable to do something different, but it needs to be done on the basis of measurement and analysis; such measurements should be considered highly contingent on the context in which they are taken, and not extrapolated to other contexts without appropriate care. ]

F u r u y a wrote on Feb 3^rd, 2009 at 10:17pm:

It is basically never appropriate to vary the NTFS "cluster size", because NTFS works on an extent basis like all high-end filesystems do.

Cluster size in NTFS only comes into play after you have stressed a filesystem by letting it become excessively over-fragmented (which is almost always due to letting it reach a critical load factor; yet another one of the negative things made more likely by over-partitioning). Under normal circumstances a properly sized NTFS partition would not benefit at all from defragmenting more than once every six months, tops.

Nigel Bree, thanks for the accurate informations. I understand your point.


Ok, so I won't change cluster size. Althought the Movie Partition are not likely to be fragmented (because of very little moving/writing files; it's basicaly read-only storage and few writing of large files).

Brian wrote on Feb 4^th, 2009 at 1:36am:

That's one of the two critical factors, because the higher the load factor in the partition the more constrained the filesystem is in choosing where to place things. For pretty much any filesystem, from 80% upwards fragmentation becomes basically inevitable and the longer things stay at those higher load factors the worse and worse it gets. That's an inevitable mathematical result, as I've explained in the past; it's a packing problem - i.e., NP hard, but made in effect impossible because the packer - the OS - is presented with items in no particular order and they can change size after being placed. More free space headroom means less likelihood of the OS initially placing things such that a later change can't be accomodated by extending the object in-place.

[ Of course, things like System Restore can cause problems by driving what appears to be a safely sized disk - relative to the actual data in use - into being critically full. Having System Restore on is the right default for consumers; its space consumption can have negative performance impacts, though. ]

The other critical factor is usage and application behaviour; the number of files being written (which the filesystem has to choose places for), and how the applications write them. The Windows OS (like most others) has APIs which allow applications to indicate how big a new file will become, so it can efficiently allocate the necessary space in a single contiguous extent; applications that don't use them (which is most of them) force the OS to make guesses when choosing where to place things, so things work less well, and in rare cases there are pathological application behaviours which defeat the OS heuristics and trick it into placing things poorly (forcing fragmentation even on a modestly loaded disk). These aren't problems in the OS, they are problems with badly-written applications.

Quote:


I agree. At one time I had multiple partitions for different files say... documents and pictures, videos, music, ghost images. I soon found that some were filling up while I still had lots of space on others. I have found that using one large partition for all data storage on a disk and seperating using folders is much easier. The only seperate partition I utilise is for the operating system. Though this is all for storage concerns and not performace.

ml3 wrote on Feb 12^th, 2009 at 12:11am:

I wouldn't say my view is the consensus at all; it's more a matter of how you want to deal with your computer, and deciding whether the negatives of partitioning outweigh the benefits. Just bear those negatives in mind, that's all, and consider that maybe some of the positives aren't as compelling as they used to be - bearing especially in mind how technology has changed, so that most people now buy laptops with only a single drive bay, for instance, and how easy it is to get enormous external drives we can walk around (unlike the hideous, unreliable optical media most folks were using for backups before), and how drives have gotten smarter and better at handling errors even as they have gotten bigger and faster and cheaper.

And other changes: software has gotten better too; FAT32 is a fading memory on most systems (which is a good thing) and NTFS is incredibly robust. Windows XPSP2 and above are in practice incredibly reliable compared to any version of Win9X that ever was, so it's actually extremely rare for people's OS to simply eat itself and leave an unbootable mess the way Win9X used to. Backup speeds really mattered when Rad was writing his guide, when you had to sit and wait in front of the machine to change media or take the machine offline while it was going on. Now, with external drives and with volume snapshot giving online backups while you keep using the machine, it's less compelling to micromanage that stuff.

The other thing you need to do in chopping up your disk is that you suddenly need to classify all this stuff to figure out where it goes. Now, that's fine if you have, for instance, a clear set of categories in mind for everything so you can figure out where they ought to go in terms of "these don't need to be backed up, this does". Figuring out those kinds of things for everything you put into your computer is, realistically, a lot of work, and the question is, is it worth doing that work?

For me, the answer is a clear no.

ml3 wrote on Feb 12^th, 2009 at 12:11am:

I don't keep the OS separate either - haven't done in years.

ml3 wrote on Feb 12^th, 2009 at 12:11am:

It's a absolutely fine idea to have a second disk to store your backups of the primary onto. Really, really good.

That said though, there's less to separate internal and external hard disks nowadays - a good drive in a quality eSATA/FireWire800/USB2 external enclosure is going to be just as quick as an internal one if you can use eSATA or Firewire800 - and to be able to just pick up and keep running when the primary drive fails is hard unless you just RAID-1 them - unless you think about having a system where you can mount and run the backup in a virtual machine if you had to, which is a *great* place to be in but again requires some planning.