Re: Hibernation
On Fri, 19 Nov 1999, Heather wrote:
>> The point is that a scrap of swap in ram decreases main memory and makes the
>> swap more likely to be used.
>>>[...]
>> The issue is that it will be used directly if it's not forced to be in swap
>> because of main memory being used by a RAM disk (for swap).
>
>1 Mb less is going to make that much difference? I detect more need for
>kernel work then.
It's an issue of the working set. If the working set for your applications
adds up to 96M and you only have 95M then performance will be lacking. This
is CS theory and not something that can be fixed with kernel coding.
Good kernel coding and applications coding can reduce the working set, but
there will be more things you want to run which will have a greater working
set.
>> Yes. However any program that gets swapped out will have it's code pages
>> discarded. So when it starts to be used again the code pages will have to be
>> brought in from disk and the drive will spin up anyway.
>> The only exception to this is certain types of binaries in the "a.out" family
>> which behaved like DOS .COM files (code and data in the same memory area).
>
>Thus a stronger need to simply not run junk you aren't using.
When it gets paged out it doesn't cause much problem. It's only when you
have dozens of megs of paged out data that performance can decrease (more
seeks in the swap space).
>> >Isn't Neomagic supported directly in the latest X? Why use a non-optimal
>> >server?
>>
>> I always had big problems having two X servers running on the same display
>> before I went to fbdev. I have VT5 and VT6 running seperate X sessions.
>
>Hmm. Glad I don't have Neomagic then.
I don't think it's any worse than other chipsets. Once I get the drivers
written it'll really rock!
>> >I have an IDE adapter cable, and it's possible to get a SCSI one, but that
>> >wouldn't really do what you want. The Omnibook 800 - as far as I can tell
>> >unique among laptops - has its internal SCSI bus accessible as an external
>> >port, so you should be able to hang a whole chain off of it. Helluva docking
>> >station you get there :)
>>
>> External SCSI requires power down to disconnect. Even with hibernation it
>> would be a PITA.
>
>PCMCIA SCSI only requires the cardbus stuff. Still have to actively mount/
>dismount though. Carrying that stuff is more the PITA.
Yes. That's why laptops should be made with more space. The IBM 700 series
used to have space for a second hard drive or second battery which was a good
statrt. I think that I can put a second hard drive in my 600 series if I
have no internal floppy or CD-ROM.
>> I use my laptop to compare Linux to UNIX servers that come from all the big
>> vendors and I like Linux to win. Recently I had the opportunity to replace a
>> Sun server with 400M of RAM and SCSI RAID with a PC running Linux (128M and
>> an IDE drive). The Linux machine is significantly faster, now I just want
>> RAID-1 on it for obvious reasons!
>
>So you make good business convincing people to buy big PCs and run Linux as
>their main servers because even your laptop kicks their present one's butt? :)
Actually I don't make good business from it. Usually they ignore the
performance issues and would rather just give more money to IBM or Sun. On
the occasions when they go for Linux it doesn't result in me making more
money (so it's not good business as such).
I just evangelise Linux for fun.
>[moved]
>> No. Just buy a new laptop. I think that you should get a new one every 18
>> months.
>
>> Firstly you don't want to lug 5Kg. You want at most 3.5Kg with 1280x1024
>> display, 500M of RAM, 4 hard drives, and CD-ROM/floppy (like the Thinkpad 600
>> series where you choose which one of floppy or CD is external and which is
>> internal). This is all achievable with today's technology, and would cost
>> much less than $32000 if they made them.
>> As for Video. I am waiting for that, but expect I'll be waiting quite a
>> while.
>
>Do *not* tell me that what I want is what you want. What I want is a laptop
>to serve my mobile needs until its components finally wear out; if it also
>meets some of my desktop station needs, great, but I'm not gonna cry over it.
>
>I actually like the smaller box; if I want a bigger monitor, I bloody well
>plug it into one. But it's not weight that has me wanting smaller, it's
>the shape.
You must not carry it much then. I carry mine everywhere. On Wednesday I
was at a bar in Amsterdam when the conversation turned to Perl and CGI-BIN
(it used to be that medical Doctors were accosted at parties by people
wanting advice). So I got out my Thinkpad and started showing some code
fragments.
>When I'm thinking "quality video" I'm thinking crisp images that don't flicker
>and a vidcard that doesn't squeal or click when I change modes because I do, a
>lot. (Yeah I can hear elctronic things make noises. Sigh.) Talk about what
>you will -not- find in the marketdroid bullet list.
I agree about noise. I'd like to get machines without fans. I heard that
IBM designed a line of desktop PCs specially for the European Patent Office
(at immense expense - but EU tax-payers have lots of money) which had no fans
at all and had extra quiet hard drives.
I think that the way of the future is for all general desktop machines to
have no hard drives and to network boot (very quiet - no hard drives and less
cooling fans needed). Then you have grunty servers with RAID arrays in
another room.
Laptops still need hard drives though. I think that all machines with hard
drives should have RAID-1 and a journalled file system. This will not happen
until the Linux kernel gets changed to make Raidtools2 a standard feature
(and the bug related to journalled file systems is fixed), and until
companies like IBM start making laptops with a hard drive at each end (so if
one end gets smashed your data will be safe at the other). Also if the
laptop just gets dropped from a great height then if you have two drives
there's a significantly greater chance that data can be extracted from one of
them.
--
Electronic information tampers with your soul.
Reply to: