Re: Disk Warrior vs Norton
From: Chris Inacio (inacio+msnews_at_ece.cmu.edu)
Date: 02/02/05
- Next message: Ignatz ratzkiwatzki: "Re: Mac VPN Connection to Windows Server"
- Previous message: William Smith: "Re: Mac VPN Connection to Windows Server"
- In reply to: Rod Hagen netspace: "Re: Disk Warrior vs Norton"
- Messages sorted by: [ date ] [ thread ]
Date: Wed, 02 Feb 2005 10:56:27 -0500
Rod Hagen netspace wrote:
> On 1/2/05 12:01 AM, in article BE239671.2F88C%me@somewhere.com, "Bob
> Greenblatt" <me@somewhere.com> wrote:
>
>
>>My answer to these questions are:
>>If you're using OS 10.3.7 don't bother. You REALLY DO NOT NEED to defrag
>>your drive. The OS does this for you. Even if you did defrag it's highly
>>unlikely you'd be able to perceive of a performance difference. Leave well
>>enough alone.
>
>
> Panther does a great job of defragging files, Bob, and I agree that
> fragmentation is far less of an issue than it used to be under earlier
> versions of the OS but in doing so it ultimately creates a situation where
> free space becomes extremely fragmented.
>
> Even though its a less pressing issue than it used to be, you still need to
> keep an eye on it, and it can indeed have an effect on performance when it
> becomes excessive. http://www.executive.com/fragbook/contents.htm has some
> useful discussion of issues like free space fragmentation on a general
> level.
>
> Cheers
>
> Rod
>
>
In general, assuming you can keep your per volume (usually a single disk
on a Mac/Pc) usage below 70%, then the fragmentation should be managed
by the file system without too much trouble. Above 80% and you start to
run into issues with how the file system can schedule the disk and
allocate space.
Most modern file systems have their directory structure designed such
that it operates as a type of b-tree. If its really high performance,
then it can schedule its read/write operations possibly across multiple
drives including elevator scheduling to get even more performance.
There are tricks to play to reduce internal fragmentation within the
file system. You can merge small files together in a PC class machine
such that they can then fill up a block. But usually this is more work
than its worth. Especially with the relative cost of disks.
Logical fragmentation, even when files are fragmented may not be too
bad, performance wise; as long as there are not a TON of fragments per
file. Files that are logically not fragmented are usually physically
fragmented on the disk anyway.
Free space fragmentation can be an issue, and it can really cause some
performance degradation if it's severe; but that would indicate that you
are creating a mix of file sizes and their lifetimes are also mixed,
some short and some long. Usually, in high performance systems, it pays
to figure out what types of files get created and their lifetimes and
give them space appropriately.
But, I would imagine, for the vast majority of users, that buying a
sufficiently large hard drive and not using every last bit of it, would
be enough to keep the system humming along just fine.
The whole goal of all of this is to not have to move the read/write
heads, unless you really have to, and when you do to get a good bang for
the buck when you do move. e.g., you want to read everything under the
new head position before you move again. Elevator scheduling is
supposed to make it so that you get the big bang for the buck when you
do move; but most non-server systems aren't busy enough to make that
work well; if you get a lot of requests to process, then you can group
the physically collocated ones together, and issue them at the same time
without adding too much latency to the process --- which is what we're
trying to avoid in the first place!
As a side note, while the details of NTFS would lead me to believe it
should work really well, I've decided in practice that it does not! And
I have yet to figure out why it degrades so quickly! And having worked
with a true file systems genius, his answer was, "I don't know, stop
using it." Maybe I should go back to the old Unix practice, 1 physical
disk for swap, 1 physical disk for temp, and then other disks as
appropriate.
Fixing a disk is a totally different story; and with HFS+, that should
be relatively painless. Gotta love journaling, (when it works right ---
you should see what happens when it doesn't!)
But I would hope for most people this whole topic is a non-issue.
Unfortunately, OS X does need a little more help in this area than it
should; but still, doing so much tweaking is probably a waste of time.
Chris Inacio
- Next message: Ignatz ratzkiwatzki: "Re: Mac VPN Connection to Windows Server"
- Previous message: William Smith: "Re: Mac VPN Connection to Windows Server"
- In reply to: Rod Hagen netspace: "Re: Disk Warrior vs Norton"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
