Advertisement
Front, Kernel, Security, Distributions, Development. See your byline here on LWN.net.
Weekly Edition
Return to the Press page
| |||||||||||||
|
| |||||||||||||
Lightweight Linux, Part 1 (IBM developerWorks)
This IBM developerWorks article
is about leveraging older hardware to break the hardware/software upgrade
cycle. "Too often, modern operating system vendors treat hardware as
if it were disposable -- use it for a year and then throw it away. One
might be tempted to believe that secret backroom meetings are going on
between vendors of operating systems and computer hardware
manufacturers. New operating systems and applications demand the latest,
most powerful hardware. The newest hardware works best only with the
latest, most feature-rich software. I'm sure the churn helps someone's
bottom line, but it does nothing for mine."
(Log in to post comments)
What is done in secret. . . Posted Nov 8, 2002 19:17 UTC (Fri) by rjamestaylor (guest, #339) [Link]
IceWM [off topic] Posted Nov 8, 2002 19:41 UTC (Fri) by tjc (guest, #137) [Link] OK, this is sort of related in that it has to do with "lightweight Linux"...Some time ago (a couple months maybe) there was an article published on how to configure IceWM. It was linked by one of the Linux news sites; I thought it might have been LWN. I've been fishing around for this -- does anyone have the link? TIA, TedC
never mind... IceWM [off topic] Posted Nov 8, 2002 19:57 UTC (Fri) by tjc (guest, #137) [Link] I found it.http://www.troubleshooters.com/lpm/200209/200209.htm
Lightweight Linux, Part 1 (IBM developerWorks) Posted Nov 8, 2002 19:42 UTC (Fri) by judge (guest, #6234) [Link] Ugh, right slackware too bloated.I have a 486laptop with 100mb linux partition and 4mb of ram. Now this beast can even somewhat run win95(websites take minutes to load in opera3). However I cant believe that the author called it too bloated. Maybe the default floppies wont install in 4mb of ram, but with some help of smalllinux floppies & little swap tricks I got the slack installer off the ground and had it install slackware for me. Of course I like the author's use of uLibc, I havent had time to set that up yet.and On my laptop as soon as it boots up(with only 5 or so processes) I'm already 1mb into swap :). However, gcc works for simple c programs. Anyways to summarize 12MB = crapload of ram for linux. You can even run a stipped down X at reasonable speed.
Lightweight Linux, Part 1 (IBM developerWorks) Posted Nov 9, 2002 11:32 UTC (Sat) by onetimeonly (guest, #7620) [Link] >you can even run a stripped down X on 12 megs of RAMWell while we're on the topic of using old hardware, let's all remember the day when major FTP sites had 512 users on a unix system with the box running 256 megs ram. Back when we all had 1, maybe 4 megs ram. Yes that's right, we pay for truly anti-aliased fonts with a geometric increase in RAM usage and processing power. Anti-aliased fonts for example are a fine example of a "binary distribution" AI. The point being is that you pay for convenience. Convenience meaning you don't need to *learn* anything to use the computer. What does this mean? simple. Since you don't need to learn ANYTHING, the computer is controlling everything FOR you just like skynet in "terminator". An extreme example I know, but it serves my point well. There must be a breakeven point where we all acknowledge there is *something* that we must *learn* in order to use computers efficiently. If we could all do that, then there wouldn't be this crap floating around about ram and OS'es wasting CPU cycles doing nothing on purpose. -frl
Lightweight Linux, Part 1 (IBM developerWorks) Posted Nov 9, 2002 1:09 UTC (Sat) by iabervon (subscriber, #722) [Link] If you're going to the trouble of making stuff small, you should be able to get the whole thing on a single floppy, instead of using a boot disk and a root disk (and a supplemental disk, especially). I've put everything he's got on one floppy, without even going for BusyBox and uClibc (libc5, bash, binutils, util-linux). You could probably even fit a suitable cvs, ssh, and vi if you use uClibc and BusyBox configuration.The other advantage of a single floppy is that, when installing to the hard drive, you copy the same floppy twice (both the ramdisk, mounted on / off of the image on the floppy, and the kernel, on the filesystem you get if you mount the floppy itself). Just don't tell your friends you were pleased to save 2.88M of magnetic storage...
Lightweight Linux, Part 1 (IBM developerWorks) Posted Nov 9, 2002 11:26 UTC (Sat) by onetimeonly (guest, #7620) [Link] A quote from your (interesting) post:If you're going to the trouble of making stuff small, you should be able to get the whole thing on a single floppy, instead of using a boot disk and a root disk (and a supplemental disk, especially). I've put everything he's got on one floppy, without even going for BusyBox and uClibc (libc5, bash, binutils, util-linux). You could probably even fit a suitable cvs, ssh, and vi if you use uClibc and BusyBox configuration. My reply is as follows: "We can fall back on platitudes as much as we want, but the basic logistic solution to this problem can be expresseed without expressing by the confirmation of understandiung:" a) Such a system requires a HUGE database and the corresponding AI neccessary to create such a "single" *neccessary* (read CUSTOM) boot disk, b) There is a CERTAIN breakpoint in computing power and application depth which we all must pass (acknowledge it; don't say what it is) before we can utiliize this certain "unique boot disk". Considering a)and b) it will immediately become apparent that depending on the power of your system, the "boot disk" can be as small as, say, 50k in size, or perhaps 24k or so. A ridiculous example I know, but the point is that what *you say* makes little sense to me at all. -frl
Lightweight Linux, Part 1 (IBM developerWorks) Posted Nov 9, 2002 13:57 UTC (Sat) by nomad42 (guest, #2089) [Link] "One might be tempted to believe that secret backroom meetings are going on between vendors of operating systems and computer hardware manufacturers."You'd better believe it - or continue believing in Moore's law as a natural law. One question remains, though: what is the economic law behind it? Why double the power in 1.5 years? To keep up with more heavy-weight OSes? Or rather to keep lousily designed applications and their follow-up releases running, without imposing the costs of reeingineering on the software manufacturers? Let's hope that the physical boundaries will be reached real soon.
Lightweight Linux, Part 1 (IBM developerWorks) Posted Nov 11, 2002 16:40 UTC (Mon) by bockman (guest, #3650) [Link] Let's hope that the physical boundaries will be reached real soon.I don't know about phisical, but IMHO _practical_ boundaries are very close, if not surpassed already. Who needs 2 GHz pentiums? Who needs Itanium-powered workstations? I think that even patological gamers and 3d designers have already more CPU, RAM and storage that they can use.
OK, that was a more than a bit stretched, but you got my point: for most everyday usage, todays PC offer more power than it is needed, even for people using the worst bloatware (although, as a developer, I do not understimate the capability of my peers (and mine) to always add more bloat :-).
Lightweight Linux, Part 1 (IBM developerWorks) Posted Nov 12, 2002 10:32 UTC (Tue) by beejaybee (guest, #1581) [Link] > Who needs 2 GHz pentiums?Well, I'm running jobs that take 8-9 days each - on a 2.53 GHz P4. There will _never_ be enough raw CPU power to satisfy those of us interested in high-end mathematical research.
Lightweight Computers, Begone! Posted Nov 14, 2002 15:35 UTC (Thu) by shane (subscriber, #3335) [Link] There will _never_ be enough raw CPU power to satisfy those of us interested in high-end mathematical research.Agreed. There have been people complaining about the new-fangled computers that we don't really need for as long as I've been in computing, about 15 years. Although I suspect John Dvorak has moved on from his 20 MHz 80286 by now. Yes, people were able to do useful work on 8-bit machines with 4 Kbyte of RAM. I'm glad those days are gone. If we can get more functional, or simply more attractive, interfaces on modern machines, why not use them?
Lightweight Computers, Begone! Posted Nov 14, 2002 22:51 UTC (Thu) by lysse (guest, #3190) [Link] The argument isn't really about more attractive interfaces. (I don't know about functional - I still haven't found anything to beat bash for functionality; it rather seems to me that we're investing vast amounts more processing power for less functional interfaces.) It's about whether those attractive interfaces really need 256Mb RAM and 1GHz CPUs to wake up. Admittedly, there are fields which need vast amounts of processing power, and users in those areas should now be delighted - processors more powerful than older Crays are now available for less than the price of a holiday abroad. But if you're devoting that much CPU to mathematical problem, chances are you don't want some GUI to snaffle 20% of it before you start; and in any case, most people don't need the processing power - especially when you consider that the main bottlenecks of modern systems are RAM and hard disk speed. Consider that using bucketloads of RAM decreases performance, as cache locality is adversely affected. Consider also (and I know it's a rough metric) that the capacity of the average PSU supplied with a new system hsa doubled within 2 years - this in a time where we are coming perilously close to a global energy crisis (how many years of oil and gas are left?) Not to mention the disastrous effects of a largely needless 3-year replacement cycle, leading to most "obsolete" (in reality, just a little slow) computers ending up in landfill sites - the very worst place for them, considering how unpleasant the processes involved in manufacturing them are. Meanwhile, for the last 5 years, the cost of an entry level computer has remained static at about £500 in the UK, which has pretty much destroyed a second-hand market (it's simply not worth selling that computer for a twentieth of what you paid for it). As I said, most people don't need the processing power they have. The games market is still striving for photorealism, but you know the funny thing? People still play Tetris (which can be implemented in 256 bytes, if you're careful). People still play that old game where you enter an angle and a power for a projectile to be lobbed at your opponent (it's the latest craze to hit my office). Photorealism seems to have eclipsed absorbing gameplay as the primary aim for game development (with honourable exceptions such as Metropolis Street Racer - but hey, what's the Dreamcast's hardware spec again...?) And games are the only mass market applications that even come close to justifying the CPU cycles they consume - upgrading the CPU won't make a great deal of difference to a word processor, or a database, as the performance of these are largely dominated by other factors (for a database, memory / disk speed; for a word processor, the speed at which human fingers can hit keys... so exactly why can't Word or StarOffice keep pace with even my typing?) Hardware and software manufacturers have an obvious vested interest in keeping these trends going. But even now, they are beginning to find that their customers have stopped listening to the hype, have noticed that they can buy for £500 today what cost them £1000 years ago. They'll pay the £500 now if they fancy it; more likely, they'll wait until next year's £1000 model costs £500 and buy it then. It's already happening, and there are already responses around (eg. the VIA Epia-based PCs) that don't cost vast sums of money, and don't sound like a jet aircraft taking off when you switch them on. And it isn't as if everyone has forgotten how to make efficient software, or as if it never existed - for example, the venerable RiscOS platform offered anti-aliased fonts, in realtime, ten years ago, on hardware on a par with a 486, inside 1Mb. Yet people routinely configure swap space to supplant their 256Mb physical RAM (indeed, some OSes insist upon it!) and the mainstream of software development (even OSS development) seems to regard RAM and CPU as essentially infinite. Which of course leads to ever-slower performance where such applications hit the "no it ain't" reality check. However, I think the practical limits have been reached. I've shelled out for two current-spec systems in my life - they have both been noisy and unreliable. Fast, yes, but I don't like using them - I have considerably more faith in my 486. These days I have severe misgivings about even investing in new technology, especially when a P166 can be had for £30. I just don't see the need. But I'd much rather be able to buy a new P166 with 2Gb hard drive for the same £30, and have it in a case the size of a 3.5" floppy drive and running for several days on a few pence worth of electricity. And then we might just see software written to live within the constraints of hardware, rather than assuming that it will catch up eventually. One day it won't, and even now the consumer has tired of the catch-up game. And that is the impetus for projects such as this one. It's not about losing functionality, just about arranging it a little more responsibly. OK, I'll go away now ;-)
|
|
||||||||||||