10/23   On modern disks, where should you place swap?  At the beginning?
        at the end (like on old ide disks)?  Can we be assured that mfr's
        think "end" is the edge of the platter?  I wanted to know in general
        and in the case of "20 Gig Maxtor drive".  Also, I do infact
        reliaze that it probably doesn't matter, but theoretically where
        should it go?  Thx.
        \_ experiment -don't know either.
                      \_ no, just sick of people asking questions that they're
                         simply too lazy to take care of themselves.  The
                         answer would be: "it depends, but probably doesn't
                         matter"
        \_ i tend to place it next to partitions used for filesystems that
           I might want to resize later on.  easier to resize swap than to
           resize another filesystem to rearrange space to make /var, for
           exmaple, bigger.  on some OS's you can even stop swapping on a
           particular partition without rebooting (as on Solaris). --Jon
        \_ I put it near the beginning, but these days when main memory
           is 1 gb or greater and much much faster than disk, it doesn't
           really matter that your placement will make vm a bit faster,
           since the overhead is just so damn high.
           \_ so it makes absolutly no performance boost to put swap at
                 "optimal" placement will be unnoticable.
              the edge of the platter?
                \_ if you're swapping, you've already lost.  -tom
              \_ Even on a budget system the disk runs at 7200 rpm and
                 has a 2 or 4 mb buffer. Whatever speedup you get by
                 "optimal" placement will be unnoticable with such a
                 disk. The thing to worry about these days is getting
                 the fastest ddr ram you can find and making sure that
                 it is interleaved properly.
              source of latency anyway).  Putting the swap near the center means
              that you're more likely on average to swing the head less.
                `\_ Even the slowest RAM is 8 orders of magnitude faster
                    than the fastest disk.
                     \_ Try 5 orders of magnitude.  Your point is still
                        valid, however.
        \_ huh, isn't the disk access going to be faster at the outer
           parts of the platter? look at the math, i degree of turn on
           the inner goes thru less distance than on the outer edge.
           \_ Aren't hard disk tracks concentric and non-uniform density
              (the inner tracks have the same storage capacity as the
              outer ones)?  If that's the case, it shouldn't impact
              performance either way.
           \_ once again tho, how can you make sure that what cfdisk thinks
              is "end" is  what the mfr thinks is "outer edge of platter"
              is there a standard to these things?
           \_ That doesn't account for moving the head (which is the larger
              source of latency anyway).  Putting the swap near the center
              means that you're more likely on average to swing the head less.
              \_ this is pointless ... use the time and energy to buy more
                 memory or kill a unneeded application. if you are interested
                 in doing this to see if the difference is measureable well
                 then obviously you will have to measure for your hardware.
                 this might be an interesting way to see how good a
                 disk simulator is.              --psb
                 \_ Actually, that was kind of my point.  Making guesses this
                    deep without actually measuring the performance isn't
                    productive.
        \_ WinXP automatically place it in the beginning 1/4-1/5 of your
           partition. Not sure if that means anything.