1 |
On 28 Mar 2008, at 22:12, Alan Milnes wrote: |
2 |
> On 28/03/2008, Stroller <stroller@××××××××××××××××××.uk> wrote: |
3 |
> |
4 |
> Your note is excellent but I disagree with this bit:- |
5 |
> |
6 |
>> If the PC is still slow then check disk-space, pagefile settings |
7 |
>> ("allow the system to manage pagefile size for me", click "set") |
8 |
> |
9 |
> unless as a temporary workaround you should always have the paging |
10 |
> file set as a fixed size to avoid worsening the chronic |
11 |
> fragmentation problem on Windows. |
12 |
|
13 |
|
14 |
I'm not arguing with you, but for me it depends on the user & usage & |
15 |
stuff. Several years ago, before XP, I used to be a Windows "power |
16 |
user" - I kept my pagefile on a separate disk and set it's size |
17 |
manually. I even monitored usage in Task Manager (or was it one of |
18 |
the utilities under "Administrative Tools"?) to see what actual |
19 |
amount of swap was used, but was never convinced of the accuracy of |
20 |
the results (it seemed so little). Whilst a pagefile of fixed size on |
21 |
a separate disk may be the "best" way to configure a swap file, I |
22 |
don't think it's optimal for most users. |
23 |
|
24 |
There are times when someone else may work on the PC, and having the |
25 |
swapfile on C: is simply what they'd expect to find, if they ever |
26 |
need to mess with it. I generally like to have systems that I |
27 |
configure for my Joe Sixpack customers to generally look "normal" and |
28 |
"standard", so that there's ease of maintenance and so that |
29 |
everything just "makes sense" if anyone else (probably less |
30 |
experienced than I) works on it in the future. |
31 |
|
32 |
Take, for example, partitioning - it's quite logical and correct to |
33 |
have a separate partition for the C: drive and another, D:, for |
34 |
users' files & folders; this protects users' documents on D: if |
35 |
filesystem corruption occurs on C:, or if a Windows reinstall is |
36 |
otherwise needed. But unfortunately this configuration needs more |
37 |
maintenance in the future if one of the partitions becomes full - |
38 |
experience tells me that there's always one user in the household who |
39 |
will not understand to use D:, and that users will try uninstalling |
40 |
programs and deleting their letters to free up space, if the system |
41 |
starts complaining that the C: drive is full. I would prefer they |
42 |
call me, so that I can delete something that's REALLY consuming |
43 |
space, or resize partitions appropriately, but they often do not do |
44 |
so, and with 5% or less free space the partition gets rapidly |
45 |
fragmented and slows down considerably (to the extent that |
46 |
defragmenter may be unable to do its job). When short of disk space |
47 |
other users may right-click on the drive properties and choose |
48 |
"compress files on this drive to save space" - this slows down the |
49 |
system even more! |
50 |
|
51 |
But I admit that - if the system has two drives installed already - |
52 |
then putting the swapfile on the second drive is probably less of a |
53 |
problem than my partitioning example. (Although, having said that, if |
54 |
this user _does_ choose to have a D: drive and intends to use it for |
55 |
something, then a pagefile.sys scattered amongst their music or video |
56 |
files might be confusing, or simply considered clutter). |
57 |
|
58 |
Just because you set the swapfile to a fixed size doesn't mean it's |
59 |
not fragmented - admittedly, if you do set it to a fixed size, then |
60 |
boot from another disk and defrag the drive then the pagefile should |
61 |
never fragment in the future, but I'm not convinced of the cost- |
62 |
benefit of doing so. A fragmented swapfile is only going to be a |
63 |
problem (I think - please correct me if I'm wrong) if the system is |
64 |
writing out a page of memory that spans multiple fragments. If the |
65 |
swapfile is contained in only (for example) two fragments then how |
66 |
often will this occur? I have no idea - and one of the reasons I gave |
67 |
up Windows on my own machines is its the sort of thing that's |
68 |
completely undocumented - but I'll bet it's not too often. |
69 |
|
70 |
A swapfile of a fixed size is a compromise between consumption of |
71 |
disk-space and the risk of running out of pagefile. I have customers |
72 |
I don't see for two years, so what seems perfectly adequate for a |
73 |
swapfile now may seem silly small when I next see them. Although I |
74 |
don't tend to monitor swapfile sizes & usage, Windows memory |
75 |
requirements have bloomed in that time - 2 years ago one might've |
76 |
gotten away with 256megs of RAM, but I'm certainly recommending at |
77 |
least 768meg now. |
78 |
|
79 |
Considering the size of hard-drives these days I guess I'm being |
80 |
silly in not simply allocating a fixed-size swapfile of 2gig (or even |
81 |
4!) and trusting that that'll be adequate for the life of the |
82 |
machine, but I don't like to waste space unnecessarily, and I'd just |
83 |
far rather the machine said "out of virtual memory, increasing swap |
84 |
file size" if it needs it. |
85 |
|
86 |
To generalise, I have two kinds of customers - those who fragment |
87 |
once a month, and those who never do. I don't think the slight |
88 |
penalty of a fragmented swapfile is noticeable to either category. |
89 |
Either their machine is quick enough, anyway, or it tends to |
90 |
horrendous slowness. The risk / hassle of running out of swap space |
91 |
is more considerable, IMO. |
92 |
|
93 |
Like I say, I'm not saying you shouldn't set the swapfile to a fixed |
94 |
size, I'm just saying it's horses-for-courses. I guess I'd recommend |
95 |
setting the swapfile to a fixed size to readers of this list, whereas |
96 |
I wouldn't to most of my customers. |
97 |
|
98 |
Stroller. |
99 |
-- |
100 |
gentoo-user@l.g.o mailing list |