[mso] Re: Defragging and outlook.pst
- From: "Peter Strisik" <lists@xxxxxxxxxxx>
- To: <mso@xxxxxxxxxxxxx>
- Date: Wed, 18 Sep 2002 13:53:28 -0800
Greg,
Great lesson! I guess I hooked you with the "drive geometry" line :-)
I am running XP and didn't know that NFTS fragments easier. So I will
defrag more often. I also have diskeeper lite, which I've heard does a
better job than speedisk, but was wanting to play with the settings that
speedisk offers.
Also, keeping files together (and keeping this thread somewhere in the
vicinity of on topic) means that I should specify all of the outlook
stuff for placement (not just the pst file) if I want to get outlook
performing better.
I really appreciate the time you took here.
..........Peter
>
> Your real solution is to defrag a little more often. The
> fragmentation of the file has more to do with how large the
> file is and how often it is fragmented rather than where it
> is stored. Optimization for localization to a specific region
> on the disk is pretty much an esoteric consideration for the
> problem you describe. Note that cleaning the file of
> aged/unwanted goodies and then compacting it occasionally
> will help reduce the size of the file and makes the need for
> writing it in non-contiguous ways a little less frequent.
> However, the existance of the fragments indicate the file is
> growing on an as needed basis and really has no unused space
> to speak of. XP is the best at this so far as the PST tends
> to clean up nicely without help when you delete a lot of messages.
>
> If you're using speeddisk, it sounds as if you might be on a
> non-9x OS (okay, written another way, it sounds as if your OS
> has an NT kernel and file system). If that's so, be aware
> that NTFS fragments faster than FAT and does take a
> significant performance hit in the early stages of
> fragmentation. However, once you get past 10% fragmentation
> on this file system, performance degradation occurs at a much
> lower rate. So your best investment is simply to regularly
> defrag the drive on a more regular basis. Once every week or
> two should be sufficient.
>
> I rarely worry about the location of files on the disk based
> on drive geometry. My preference, where possible, is to keep
> files that are used together stored together. That method
> allows more time to be spent reading the data from disk
> instead of losing time to moving the heads across the
> platter. There's also this to consider (and it's tough to explain);
>
> All sectors on a disk are physically the same size. So, on a
> track that is closer to the spindle/hub, fewer data segments
> will be found than the farthest track from the hub. Since all
> sectors are the same size and the outside track of the disk
> has a larger diameter, this only makes sense. Therefore, if
> you want fewer head movements to read the data of a large
> file, move this file to the outside of the disk. If the files
> are small, though, you can place them wherever you want but
> its better to try to save that outside sector for those
> really large files.
>
> If you're wondering why all those sectors are the same size
> and how this wild series of questions about where to store a
> file come from, consider this example. Draw a large circle on
> a piece of paper. For convenience's sake, make the circle
> about as large as the bottom of a standard coffee mug. In the
> center of the circle, draw a circle about the size of a
> quarter and fill it in. Yes, you may use a crayon if that
> makes this a fun exercise for you.<g> The little circle
> represents the spindle or hub of your disk platter (most hard
> disks have 3-4 platters). The big circle is the outside edge
> of the platter disk.
>
> Now, starting from the spindle area, draw another circle
> enclosing the hub that is about a quarter of an inch larger
> than the hub. Make another circle enclosing the circle you
> just made. Keep on doing this until your last circle is about
> one quarter inch smaller than the first big circle you drew.
> We'll call these tracks for the time being. Before you do
> anything else, repeat this same procedure on another piece of
> paper so that you now have two pictures of a disk platter.
>
> Okay, with the first picture, draw lines across the disk as
> if it were a pie that you were slicing into 8 pieces. Now you
> should have 8 pie shaped pieces of disk, each of them broken
> into little bands about a quarter inch wide. Those toward the
> big end of the pie slice are longer than the ones closer to
> the center, right? If each of these were to be considered a
> sector, it's obvious that some are larger than others and
> that they can't all hold the same amount of information if
> storage is based on how much room you have (if you have an
> attic or a basement then you already know without a doubt
> that storage space is a function of size, no matter how you
> slice the pie<g>). Obviously, this is not an efficient way to
> divide the storage.
>
> So with that second picture of a disk, do this: Start from
> the hub of the disk and drawa line between the hub and the
> next circle. Measure about an inch from that line and draw
> another line. Do this until you've divided that inner circle
> into several similar sized pieces. At the last line you drew,
> extend the line to connect it to the next circle out. Make
> this longer line much thicker than the rest of your lines so
> that you can see it more easily than the rest of the lines
> you've drawn. In that new ring, again, draw lines about an
> inch apart. When you've made it all the way around, again,
> draw a thicker line that ties to the next outermost ring.
> Repeat this entire procedure until you have made your way all
> the way around the outside ring with these little lines.
>
> Now you have a picture of how data is actually stored on your
> disk. Each of the sectors are about the same size and each
> line that extends from one ring across to the next represents
> the end of one track and the beginning of another.
>
> Get the idea? This makes it really easy to tell how much a
> disk will hold, too. The formula is based on how many heads
> (or readable platter
> surfaces) multiplied by how many cylinders (tracks)
> multiplied by how many sectors per track (or cylinder)
> multiplied by how many bytes can be stored in each sector. So
> a disk with 16 heads, 1024 cylinders, 63 sectors per track
> and 512 bytes per sector holds 528,482,304 bytes or
> (528482304/(1024*1024))= 504 megabytes. BTW, this formula was
> also the standard specified in the original IDE drive spec
> provided by Western Digital in WD1003. That's right, the
> original IDE specification only allowed for a hard disk to be
> as large as 504 megs. I bought my first drive of this size in
> 1994 and paid $320 US for it. Times have certainly
> changed...but the math and the logic behind it still hold!
>
> I know, this is much more than you wanted to think about but
> hang on to it. You may find it helpful when you try to answer
> a question like this for yourself in the future.
>
> Greg Chapman
> http://www.mousetrax.com
> "Counting in binary is as easy as 01, 10, 11!
> With thinking this clear, is coding really a good idea?"
*************************************************************
You are receiving this mail because you subscribed to mso@xxxxxxxxxxxxx or
MicrosoftOffice@xxxxxxxxxxxxxxxx
To send mail to the group, simply address it to mso@xxxxxxxxxxxxx
To Unsubscribe from this group, send an email to
mso-request@xxxxxxxxxxxxx?Subject=unsubscribe
Or, visit the group's homepage and use the dropdown menu. This will also allow
you to change your email settings to digest or vacation (no mail).
//www.freelists.org/webpage/mso
To be able to use the files section for sharing files with the group, send a
request to mso-moderators@xxxxxxxxxxxxx and you will be sent an invitation with
instructions. Once you are a member of the files group, you can go here to
upload/download files:
http://www.smartgroups.com/vault/msofiles
*************************************************************
Other related posts:
