Slashdot Mirror
Booting A PIII System In .8 Seconds
gizmo_mathboy writes: "General Software has announced the fastest BIOS boot time on record. The embedded system was clocked at 0.8 seconds from system power-on to transfer of control to LILO. This was on an Intel SOYO motherboard (440BX chipset) running a PIII 400. I think the quote of the article is: 'This Embedded BIOS quick-boot operation allows the device to restart and resume operations well within three seconds -- the maximum amount of downtime allowed per year for a device that must support "seven nines" or 99.99999 percent uptime.'"
sometimes i would like to have a sec more to press the delete button to get into the bios and change a few settings.
I've been waiting for something like this for a while! My car MP3 player takes too long to boot up... can't wait to get my hands on this. No mention of cost, but I've sent an e-mail to their contact link and will reply to this message with price if/when they get back to me.
I am disrespectful to dirt! Can you see that I am serious?!
Was that supposed to be 1400 MHz PIII, or PII/400, or some other speed of PIII? (Or was it a Celeron, or 286?) I checked the article, they got it wrong there-- not /.'s fault.
Dear Slashdot: next time you want to mess with the site, add a rich-text editor for comments.
It's taken 20 years, but they finally got a system
that can boot faster than a TRS-80.
Seriously
The SPARC architecture does not use a BIOS with REAL mode drivers for booting. It has protected (or whatever it is in non-x86 parlance) mode drivers built right into the firmware. On x86, the BIOS contains Real mode drivers, THis was fine for operatings systems like DOS and Win 3.1. However, modern OSs (Windows, Linux, etc) need protected mode drivers. BY placing these right in the fimware Sun is able to smoke x86 performancewise ALWAYS. I thin its time to ditch our legacy DOS hardware and start getting x86 machines with protected mode BIOS drivers. Anyone with more technical information, please comment.
-zr
OK, while that's a pretty nice thing, what's the big deal? That .8 sec is only [button push] to the lilo prompt. So? The box STILL has to boot. What if you've got a box that still has to fire up a bunch of daemons before it's even online and usable? What if it was a dirty shutdown (and the silly fool is still running ext2) with a nGig drive(s)? How does this help uptime?
I'm not a prophet or a stone-age man,
I'm just a mortal with potential of a super man.
Wow! Geek perfection! Cold to LILO in 0.8 seconds. Women will flock to me!
Now if I can just get LILO working again...
I would rather be ashes than dust!
Okay, for everyone alternatively complaining that this is overkill on the desktop, or that they would prefer all the checks, etc. in place... this is NOT built for desktop systems.
Read the post; this is for an embedded system requiring seven nines. Though it can (and most likely will) be adapted for desktops, any desktop running this will be a high-reliability server, with all the checks (except memory, which this chip does after a fashion) built into the hardware...
I am disrespectful to dirt! Can you see that I am serious?!
Slashdot also has seven-nine uptime. Except it's not 99.99999%, it's 9.999999%.
>I mean, seriously, what's the big deal if it's >0.8 seconds from power to LILO? I, personally, >would rather have a BIOS that takes a few >seconds to check the RAM, auto-detect devices, >and check SCSI drives before it tries to boot >the system.
It says "embedded system". If you put a cpu in as a brake or steering controller into something that moves at any reasonable speed, you would like it to return service as soon as possible after a power glitch.
No, this does not mean that they make embedded controllers with crashes as a design goal. It means they want to make something that is as error-resistant as possible. Not for your desktop box, in other words.
Obtw: very few of those systems have anything like Linux or Windows on them, even though some people would like to tell you otherwise.
The last thing I want is my Microwave running a Pent anything and some variant of Windoze. When that happens I'll be out back with some wood, rock, and flint to cook my meals :)
Top Most Bizarre/Disturbing Error Messages
(This is probably going to be flagged as a troll or flamebait, but think about it. We have put up with crap for so long, that when we finally get sub-second boot times back it's a big deal. It's like hiding toys from your kids for 6 months and then bringing them out as winter sets in - they get all excited about stuff they used to have.)
I'll admit ignorance as to all the required checks, double checks and initialization that must go on to get a decent OS up and running, but I still can't help but think that inefficiently designed / written bloat-ware could be done much better to improve the boot times of modern machines. Why not lazy load the drivers, etc as required?
I'm in my right mind and I have the answer to everything!
What I'd like to see is an EPROM to contain a kernel. Any kernel. A generic 8 megabyte EPROM that will hold the linux kernel (or, if you are stupid, a doze kernel) and various modules so that you can very quickly bring up your system instead of waiting for this daemon, that daemon, etc.
Sure, some of this stuff would still have to take place outside the EPROM, delaying powerup-to-input time but it could be minimized by sticking as much as the user wants (and can fit) into the EPROM. It should be generic so that it can hold doze, linux, sunos, freebsd, whatever you want as your primary os.
My system remains much the same from boot to boot so I wouldn't need to constantly reprogram the EPROM to fit with my latest change. There could be a simple utility like a BIOS upgrader app to handle the EPROM programming. Make it so that it isn't absolutely required so that if something goes wrong with the EPROM you can still boot off you harddrive - which you would need anyway if you use a bootloader and want to bootup another os periodically.
What is it that prevents this sort of thing? Design mobos with a new chipslot for the kernel EPROM and design the chip to contain enough mem to hold any one of the os kernels that are generally used (or likely to be used). If there is a lot of leftover space, perhaps you could fit another kernel and supporting modules into it until it is full allowing for a very fast dual boot setup.
In Bushworld, they struggle to keep church and state separate in Iraq as they increasingly merge the two in America.
Two comments:
Doesn't Soyo mean "gentle" in Japanese?
And, if you really need "7 nines" uptime, you shouldn't be relying on a single processor -- you should be relying on a processor farm that supports hot-swappable processors, so you can lose one or two or fifty and only lose a fraction of performance for as long as it takes to replace those processors.
On a high availability machine, there should never be any reason to reboot until you must upgrade the kernel, and I'm sure there are ways to do that without requiring a hard reboot. IBM has had farms like this for years.
--brian
Umm, anything like a brake and you do NOT WANT a power cycle. If that happens, you're toast. You go out of your way in any CRITICAL embedded system to ensure that you NEVER NEED TO CYCLE the power. So a 2 or 3 second boottime is negligable.
... I dunno. maybe bad example, but I can't think of much where the difference between a .8 and 3 second POST time is going to be a big deal at all.
In any serious embedded device you want to make sure that you never never never need to reboot the device. So I think that the only cases when boot-time would be critical is things that need to be active within instants after a poweron. Things like power backup systems or
If God gave us curiosity
99.99999% uptime = 3 seconds downtime per year
99.9999% uptime = 31 seconds downtime per year
99.999% uptime = 5 minutes downtime per year
99.99% uptime = 52 minutes downtime per year
99.9% uptime = 8 hours downtime per year
Here's the URL spelled out in full to get past the $&#$(! filter: http://www.intc.com/pressroom/archive/releases/dp1 02599.htm
What I may sound like and what I am are two totally different things. I've already got a 1GHz server in my house and bought a 700MHz AMD Athlon when they first came out. I tend to be an early adaopter. But even so, there are some things that , to me, are overkill. And Pentium III processors running non Real-time OSes in appliances just don't get me excited. No thanks. Embedded devices have to start fast but they also have to be super stable. In teh embedded arena - simple is best to avoid getting burned later (I do embedded design) I'm perfectly happy with my Microwave - its 10 years old and works just fine. My chicken isn't rubbery cause I don't cook in my Microwave. I have an oven, stove, and Weber grill to do that. My microwave is for reheating leftoves and well, leftovers never taste like they did when tehy were cooked.
The point here is that just because you think some things are overkill doesn't mean you have no imagination. Hell, I've designed home automation gear - and I still think the idea of networking all your appliances together is looney. Scanning items in my fridge? Yeah - that's gonna happen - talk about a waste of time! But hooking up an central controller to your house systems (HVAC So don't be so quick to judge next time. I'll probably buy the first 0.13 micron Athlon that comes out. But a Microwave running a PC based controller - gack - no thanks.
Top Most Bizarre/Disturbing Error Messages
77.7777777% uptime is better than 9.999999%
Best Slashdot Co
Over Y2K the company I worked for insisted EVERYTHING be turned off and physically unplugged from any network point, power point and phone point.
The cause for their worry wasnt a Y2K glitch in any of those sub-systems. They were worried mostly about a power surge or spike when everyone who had turned off their equiptment over Y2K turned it back on again just after midnight.
Of concern to us was an old HPUX system. We just werent sure if we could reliably turn it off and then get it going again. The reason?
It hadnt been rebooted in 10 (TEN) YEARS.
Video meliora proboque deteriora sequor - Ovidius
you think that you can fit everything that you need for a linux or windows bootup in 8MB of flash? You think that you can fit BOTH of them?
The linux kernel is small, yes, but that's because all of the needed modules and drivers aren't in it! They're loaded on the hard drive.
Not to mention that flash is very slow... and expensive...
You'd be better off to store a memory image of a booted kernel at the beginning of your hard drive, along with all necessary information to initialize all of the hardware. Just have a small bootstrap/lilo type of thing that quickly loads up enough to access the hard drive and file system, then load the rest into memory directly, then initialize the hardware.
But I reboot so infrequently that it doesn't really matter how long it takes. Hell, I have my system set to do a full memory check on bootup. It takes an extra 45 seconds, so what?
And stop bashing windows... My W2K Server has been up for 145 days now and counting. Check it out along with CodeRedII attack info realtime (yeah, shameless plug) =>
If God gave us curiosity
Then I suggest you look into this page that is working on a Linux BIOS.
Enjoy.
I'm not a prophet or a stone-age man,
I'm just a mortal with potential of a super man.
Rather, I think the issue is minimizing the unexpected downtime that occurs at a critical moment. Lets say you are using a computer in surgical equipment. Let also say that, heaven forbid, there is a bug in the software code. While the surgeon is busy fiddling around inside the patients head, the equipment freezes up. Every second that it takes to restart the equipment, there is the possibility of harming the patient.
Obviously, equipment this critical should never ever crash. It's nice to know, though, that should something happen, the equipment will restart quickly.
Come test your mettle in the world of Alter Aeon!
Seven nines of uptime is 315 seconds a year. I think you mean nine nines of uptime. Or maybe you mean seven nines after the decimal.
Hmm...
60 seconds * 60 minutes * 24 hours * 365 days = 31536000
99.99999% of that is (0.9999999 * 31536000) = 31535996.8464
31536000 - 3153996.8464 = 3.1536 seconds
Or the quicker way: (60*60*24*365 * 0.0000001 = 3.1536)
Looks like you forgot your decimal place somewhere.
You can accomplish anything you set your mind to. The impossible just takes a little longer.
99.99999% uptime per year?
;)
3.15s/year downtime.
This is possible with a reboot? of any machine? Even if the post takes 0.8 seconds?
I think the only way to have that type of uptime is to not go down at all.
Or of course, you could say that you have a 99.99999% uptime (average, over 1000 years)
If God gave us curiosity
Unfortunately lilo is set to prompt with no timeout, it sure gets there fast though.
(my emphasis)
A pet peeve of mine is that PHB's think that "device" uptime is the same as "system" uptime.
Decades ago, we had fault tolerant systems that had large-chunk redundancy. An entire mainframe could fail and the system kept serving.
OTOH, haven't you ever had a failing app take down your system, while running on perfectly healthy hardware?
The reason this misconception, that perfect-hardware==perfect-uptime, frustrates me, is that the PHB's get sold this bill of goods by hardware salespeople. Then they don't even allow for downtime to upgrade the effing OS every two years. Nor do they allow for a second system to either (a) take the load during an upgrade, or (b) test updates to the application.
For this silly reason, giant, fault-tolerant boxes are hurting, rather than helping, high-availability computing. Bosses would rather spend money on sexy hardware that won't solve the problem, instead of paying smart people who can design-in the uptime with combos of hardware, software, and procedures.
Quench-rant (for now).
....if only I could get my X-windows profile to load in .8 seconds. Ahhh...
Dude, have you looked at your graph recently? You have some negative numbers at the beginning of August. Does your attack counter program subtract from the count when your server tries to attack other servers?
If it ain't broke, you need more software.
Even though this BIOS was intended for embedded machines, fast boot is also important for desktop PCs. Consider the Canon Cat designed by Jef Raskin (see "The Humane Interface" by Raskin). It takes a very short amount of time to boot, all you have to do is start typing and the computer powers on and loads the operating system, putting the cursor in the document exactly where you left it off. Not only that, but there is a hardware buffer for the keyboard so that it doesn't even lose the keys you typed while it was booting up. Now that is a computer designed with the user in mind. I'd like to make a PC operating environment and the first thing I'd do is make sure it boots fast. I was thinking the BIOS would be the slow part but if it's possible to speed that up, then that is all the better.
On the BIOS topic, do you know where I can find BIOS that talks via ttyS0 instead of the video card?
To me, this is a requirement for servers. It means I can completely administer it remotely via a cheap terminal server. It also means I can dump the video card in my servers. Sure, Linux can use the serial port for its console, but that doesn't help me when Linux isn't booting. I know there are remote KVMs, but they aren't the cheapest thing in the world.
You can *easily* fit all drivers to boot almost any given system in 8MB. In fact, you can almost certainly do it in one meg. What you can't do is have all the drivers in existence availabe -- but who needs that?
Let's get out our calculators, class...
365.24 days per year (from Space.com. I don't know of any more accurate (more decimal places) numbers than this. Even if you were to add 5 whole days to the year though, it won't even add one one-hundredth of a second to the final result, so I think we can go with this).
99.9999999% of 365.24 = 365.23999963476
365.24 - 365.23999963476 = 0.00000036524 days
This is the maximum allowed downtime.
Assuming a day is exactly 24 hours long (I'm fairly sure it is),
0.00000036524 days = 0.00000876576 hours
0.00000876576 hours = 0.0005259456 minutes
0.0005259456 minutes = 0.031556736 seconds.
Thus, 99.9999999% percent uptime requires NO MORE than ~0.0315, that's three hundredTHs of one second, downtime, per year.
Nope, don't think we're there yet, but you keep pushing that 99.9999999 number if it makes you look good. After all, the general public can't do math either... =)
-Kasreyn
Kasreyn: Cheerfully playing the part of Devil's Advocate to hairtrigger
funny =)
Umm, I just ran it and it reported fine, no negative numbers.
Maybe it can't hold more than one person accessing it at once =). It wasn't exactly written to be robust and hold load. I dunno, I figured that it wouldn't be around for more than a few days so I didn't exactly code perfect...
If God gave us curiosity
That means, if you add the 75 seconds for windows to boot, you get... oh. about same as it was.
If you can't see this, click here to enable sigs.
Here, I took a screenshot of it: here. I think it might just render different on Mozilla.
If it ain't broke, you need more software.
I mean, seriously, what's the big deal if it's 0.8 seconds from power to LILO? I, personally, would rather have a BIOS that takes a few seconds to check the RAM, auto-detect devices, and check SCSI drives before it tries to boot the system.
Now realistically, does a processor even need 0.8 seconds to run a memory test and scan a few devices? 0.8 seconds is an eternity on any processor made in the last ten years.
News, hmmp, check out MRBIOS. I first discovered them back in 92/93. Back then they had auto IDE detection, support for big IDE drives, and of course a FAST boot option. A year or so after that they had (software like the promise) IDE RAID support in the BIOS. Today I still have a VLB 486 machine (my firewall/webserver) with MRBIOS. It has a 60 Gig harddrive and a 16.7 Gig harddrive plugged into it with a massive uptime ratio (greater than a year and a half at up at one point). The machine is sub .8 second warm reset times. Its basically instant. The screen clears and lilo starts booting linux (I have lilo configured not to stop unless the shift key is held down) if I press the reset button. If the machine is comming up from a cold start the bios flashes post for something less than a second and then displays a flashing "Waiting for Harddrive to spin up" while the harddrives are going Whhhhhhhhmmmmmmm... As soon as they sound spun up the machine starts booting. I have machines that are 3 years old that don't support 16 gig drives and this little box is getting on towards 10 years old and it has a 60 gig HD plugged into it. I put a dx4-100 overdrive in it a few months back and the board which was bought right when the Dx2-50's first appeared. Poped up and said,"Newer than Dx2 486 at 99mhz" ,or something like that.
Its really sad though that these guys never caught on. Most of the 'cool' bios features that have appeared in the last few years in award/AMI were in MRBIOS in the early 90's. Now they are just a shell of a compay and they don't have BIOS's for machines newer than a few years old. Some people are just ahead of their time, Well I guess i'm going to go home and reboot my machine from the third harddrive now... lol..
The idea that someone today would actually _want_ a 16-bit legacy BIOS in a newly designed embedded device is laughable. The only reason to do it is if you want to run embedded DOS (gag!) or Windows, which ain't likely if you're shooting for 99.99999 percent uptime.
Now I use a variation of Linuxbios. It works great, it's free, and I'm free from debugging somebody else's crappy assembler.
314-15-9265
LinuxBIOS is at least that fast and is GPL. It's also been around for at least a year now.
This BOIS is a building block to be used in a system that needs to power-up quickly. An example of a place it might be used is a set-top box. People don't want to wait 20 seconds for their TV to come on, and people have also become a lot more aware of appliances that don't really shut down when you turn them off. I can see a market for this BIOS, but I don't really see it in the high availability market.
In my limited experience with a telecom product that needed 5 9s uptime, everything had a level of redundancy, because you had to assume some hardware was going to fail. That means that you are possibly running at reduced performance while the system is comming back online, but the system doesn't go down just because part of it was rebooted. The 79s thing sounds like something someone in marketing though sounded good, even if it's not that applicable. It has that cool buzzword, marketing feel to it.
I thought they meant 7 nines after the decimal place. =P
Hehe.
-Kasreyn
Kasreyn: Cheerfully playing the part of Devil's Advocate to hairtrigger
Who was comparing clock for clock? I was comparing SPEC benchmarks.
Oh, and you are wrong about clock-for-clock too.
500MHz UltraSPARC-IIe (Blade 100) SPECInt 165.
1400MHz Athlon SPECInt 495
The UltraSPARC is SLOWER clock for clock than the Athlon!!!
Fear: When you see B8 00 4C CD 21 and know what it means
Ahh, the memories.
Education is a better safeguard of liberty than a standing army.
Edward Everett (1794 - 1865)
You can definately do it in 16 megs since I've seen a diskless linux box run on just 16 megs of ram. What I'd love to see is a barebones vmware-like OS done in ROM. I'm using windows on vmware on linux now (in raw disk mode) and I forget sometimes that I'm not windows directly, it's that fast (linux on windows sucks however). Of course my stupid BIOS doesn't let me stop the memory check so it takes me god-knows-how-long to check my half gig of memory (vmware loves memory).
With a vmware-like device you could easily store a memory dump to disk after bootup, and then load that directly into memory every time you restart (unless you need to update drivers).
ok then your [sic] infringing on my copyright! Could you as [sic] me next time before STEALING my comments for your own?
Desktop operating systems are designed for configurability, not zippy boots for HA. It has reams of code that's only there to make reconfiguration easier. You can pre-set a desktop kernel with its fixed peripheral configurations, but there are embedded OSes designed for that.
Linux would only be repurposed to do this sort of thing because it's open source, and because hackers understand its operation. That doesn't make it the right tool for the job.
--Blair
I think is goal was to simply give an extreme example. No kidding you don't want this type of embedded system in a brake control system.
aah ok, the problem isn't negative numbers, the problem is that you browser doesn't render DIVs correctly... I use a div for the graph with a height setting corrosponding to how high the graph should be (concept! :>) .. I had this problem on mozilla to, it has a minimum height apparently, and if you specify less than that height it displays the box much bigger than necessary. Try to load it with IE or netscape (dunno about the latter, but I know that it works in IE). Thanks for the feedback thou =)
If God gave us curiosity
My VIC-20 will also boot up in .8 seconds, and it only has a 1-MHz processor, and it'll go all the way up into the command-input mode!
READY.
with memory being DIRT FSCKING CHEAP... half gig isn't that out-of-this-world anymore, eh?? =)
Umm, yes, a totally stripped down and trimmed version of linux could run in 16MB, but a fully fledged and usable version? I doubt if it could fit in 16MB. Do a memory dump sometime to see how much memory your system is using right after bootup, I'd be interested to see actually.
You can't just do that load-a-memory-image trick because all of the devices in the system need to be in the exact same state that the drivers think that they are in. Thus if you were to just load a memory image, all of the drivers in the system must support hot-reset, including kernel drivers. Windows 2000 finally supports that (hibernate mode), so you could hypothetically boot up, immediately hibernate, and then just always use that hibernate restore. that would speed things up considerably. My system uses about 70MB total on boot (but I'm sure that if they were smart and didn't load stuff that "might be used soon" it could be 40MB. Compressed that's easily 30MB which is just a second or 2 to load off the disk. That's gotta be faster than doing a full boot all the time.
Anyways, it's not just as easy as it sounds. I don't think linux has it yet, does it? Is it in the works? Do linux users even want it?
I don't know why they don't implement that as a standard feature, but I'm sure they have their reasons.
If God gave us curiosity
OK, I stand corrected!! =)
But seriously then, what's the speed difference between that and just loading the image off the disk?
If God gave us curiosity
Am I the only one here who remembers DOS3.3 machines that had DOS in ROM? By the time you took your finger off the power button, you were at the prompt, or your autoexec had run.
-fb Everything not expressly forbidden is now mandatory.
yeah, Windows XP monitors file access on bootup and rearranges files loaded on bootup in the same order that they're loaded on the hard drive, so after a few boots your system actually boots quite a bit faster, it's impressive. One of those "Duh, why didn't we think of that before??" =)
Compared to your boot process there, what's the difference between that and just a
* Spin disk at power on
* load simple BIOS that loads and uncompresses first xMB off the HD into ram
* execute from extracted.
I don't think that'd be much slower (if any slower at all, since you have to do the disk read anyways), but this way you don't have to worry about the whole BIOS thing.
Here's another silly question. Can't you postpone most of the BIOS checks to happen after the kernel has started loading? I.e. once you've decompressed the root kernel image from the HD into memory and the base kernel is loaded, do some of the less-important BIOS checks concurrently then (i.e. upper range memory check, floppy seek, etc, etc).
If God gave us curiosity
> with memory being DIRT FSCKING CHEAP... half
>gig isn't that out-of-this-world anymore, eh??
NVRam is not "dirt fscking" or any other kind of cheap. Also, RAM for embedded devices is not always cheap like it is for consumer devices.
Since we're talking about embedded devices, I think it's fair to point this out.
>Umm, yes, a totally stripped down and trimmed
>version of linux could run in 16MB, but a fully
>fledged and usable version?
It depends on what you mean by "usable"
I have a notebook I use all the time with only 16MB. It's nothing like "totally" stripped down.
Trimmed, as you'd do for any notebook install, but not as much as you seem to think. The same notebook runs windows95 just fine too.
-fb Everything not expressly forbidden is now mandatory.
>0.8 seconds is an eternity on any processor
>made in the last ten years.
On the *processor*, yes, but not on memory chips, address decoders, rom's, the things that are being checked and read during POST. Those devices are still slow.
To get a very good indication of how slow your memory is, try this memory diagnostic (x86 only):
http://www.teresaudio.com/memtest86/
Not only will this debug certain classes of problems that are otherwise hard to track down, but it will make you realize that the fast processor is not the whole story.
-fb Everything not expressly forbidden is now mandatory.
The major problem is the chipsets and the 20-year-old designs they're based around. Drop in full 32-or-more-bit DMA controllers or require all peripherals be bus-master capable, segregate the ISA bus to its own out-of-the-way 16MB window somewhere (see: Apollo DNx000 family), hardwire a handful of interrupts and a hardcoded address range to each slot (see: EISA), drop the legacy keyboard/mouse interface, and redo IDE entirely (see: SCSI). While we're at it, let's scrap BIOS and replace it with OpenBoot. Now there's a machine free of legacy crap that might be worth writing home about.
Geewiz, that sounds like my Mac (G3).
But you're right, there is a need for a chipset redesign, not a processor redesign (at least for now). Thankfully, Intel has the Itanium. It's a start, at the least.
My question is, what do IBM's Power-based systems use to boot?
NVRam is not "dirt fscking" or any other kind of cheap. Also, RAM for embedded devices is not always cheap like it is for consumer devices.
The comment I was replying to was that the person had a half gig in their home machine, and since 128MB PC133 is $18.95 at the corner store now, fully populating your motherboard isn't expensive AT ALL. I can't believe how fast and quickly memory prices have fallen!! Can anyone else remember when it was like $30/MB back with SIMMs? Then with the DIMM price hike? Damn... So yes, SDRAM is dirt cheap right now =) Kinda pisses me off. About a year ago I paid $400 for 256MB RDRAM off ebay for my P4. Now it's like $81 for the same amount.
I have a notebook I use all the time with only 16MB. It's nothing like "totally" stripped down.
Trimmed, as you'd do for any notebook install, but not as much as you seem to think. The same notebook runs windows95 just fine too.
And what do you run on it? No, seriously. I want to know what you can run on a 16MB laptop, deemed "usable". If you're referring to maybe wordpad or 1 small old application, ok sure. But I remember with 32MB on a laptop trying to run Word with Win95 and it choked bigtime. Took forever to load, was ok to use provided that I didn't try to use any other application at the same time.
I think the point of this entire thread was that they were talking about doing this to your normal home machine to speed up booting time. I'm just not certain that it wouldn't be a lot easier, and no slower, to just put that block at the beginning of the hard drive and load it from a very simple bios. This way you're unlimited as to what you can do, you don't have to flash to test anything, and since you have to read stuff off disk anyways sooner-or-later during the bootprocess, do it all at once at the beginning and grab your bootstrap/ramimage at the same time.
If God gave us curiosity
Actually, both this solution and LinuxBIOS can get to a command prompt (or, more likely, load an embedded application) in 3 seconds.
The 0.8s to LILO prompt is nice because it demonstrates how fast the BIOS is, but is irrelevant from a system designer standpoint: The important thing is how fast they can get their program running.
4 year uptimes for NetWare are not that uncommon, especially in a controlled environment where they are assured of consistant power.
load "linux",8,1
Umm, yes, a totally stripped down and trimmed version of linux could run in 16MB, but a fully fledged and usable version? I doubt if it could fit in 16MB.
The idea is to put a minimal kernel (no apps) in the 16MB, and then DLKM the rest. Actually, my idea was to put a minimal linux with just enough support to run freemware or some other vmware-like product, directly in the kernel. I have no idea how much space that would take, though. It's basically a microkernel architecture, the meat of the OS (filesystems, networking support, etc) would go on top of it. Besides booting quickly, you'd hardly ever need to reboot in the first place, because the majority of the code will be dynamically loadable.
You can't just do that load-a-memory-image trick because all of the devices in the system need to be in the exact same state that the drivers think that they are in.
I did this all the time with vmware. I suspend the linux to disk, reboot windows, then resume. The TCP gets a little messed up, but it's quite simple to just bring the interface down and then up again. Pending disk I/O is presumably flushed before halting the OS execution, I'm not sure exactly how they do it but they do, and I'd imagine most of their work is spent making it compatible with guest operating systems, if you had the actual support of the guest operating system it'd be a lot easier.
The only potential problem I see for this approach is with gaming. Perhaps it could be worked around with raw I/O access and direct screen writes, or maybe you'd have to modify the intel architecture itself, I'm not sure. But other than gaming, introducing a small amount of latency into system calls is worth it IMHO for the gain in reliability. Again, I'm running windows on vmware on linux and I am noticing zero problems. This on a 500Mhz celeron with half a gig of memory (256 megs dedicated to windows). An earlier article mentioned about how people buy machines which are way too powerful to be used, I think we've reached the point where we can start trading processing speed for features.
ok then your [sic] infringing on my copyright! Could you as [sic] me next time before STEALING my comments for your own?
Don't get me wrong, this is great, but is it worth the implementation? The numbers say no.
I think one of the big turn-offs people have about comptuers is how long they take to boot up. Computers will be more popular with the "teeming masses" when it's something they can flick on like a TV.
And there's nothing wrong with that mentality either. Computers are tools. My machine takes like a minute to boot the OS... how many other tools or appliances take that long to boot? What if your TV took 60 seconds before it could perform its task? Or the phone, etc. Even cars don't have to be warmed up for that long in average conditions.
Of course, true geeks just leave their boxes running 24/7 anyway so boot times aren't terribly important.
OtakuBooty.com: Smart, funny, sexy nerds.
I think it is possible, but it isn't going to look anything like a PC. I'm used to seeing high-reliability systems specified as having 99.999% (5 9s) availability. You can tack on more 9s, but you are going to have a hard time proving that the system is actually that reliable. There are straightforward techniques for estimating hardware reliability. Estimating software reliability is much uglier. Statistical methods can be used to create system test plans and to interpret the failure data. The cost of the testing increases very quickly as the desired reliability increases.
Perhaps you missed, or chose to ignore, the fact that the links you posted to the CVSweb filter are calling files out of the Attic.
The Attic is for deleted files.
So, bitchslap and modslap aren't being used anymore. Yeah, the spirit of consorship probably lives on in other pieces of /. but you do your cause a disservice by pulling up deleted code and treating it like it's live.
You cannot apply a technological solution to a sociological problem. (Edwards' Law)
I maintain that OS'es should save as much system state across power-downs as they can, along the lines of APM sleep/wake (or better yet using an OS built out of persistent objects that can boot instantly and page in whatever is needed, on demand). Hell, with that no-POSTing BIOS and APM sleep/wake, you can already do a ten-second power-on without harming any non-volatile memory devices whatsoever. THIS IS NOT ROCKET SURGERY, FFOLKES.
-jhp
/. -- the Free Republic of technology.
Now that we've got a PIII booting in 0.8 seconds, to achieve "seven nines", we have 2.2 seconds spare. What can we do with this time? I'm sure we can do a lot of valuable system maintenanace in this time that we would not have otherwise been able to do.
We could:
Of course, you might have to work fast....
The only thing necessary for the triumph of evil is for good men to do nothing. - Edmund Burke
Why are the geeks so concerned about things being as fast as possible? What's an extra 5 seconds of your life?
Because every bit of performance does matter.
Case in point.
I'm not talking about time you could go get coffee while the BIOS does POST. Take that time saved to reboot, and multiply it by the number of reboots. Now suppose the blue screen code simply jumped directly into the reboot code! Like magic! A system that never blue screens! It is always up. That's innovation!
And this is not the first instance of a major advance in the evolution of computers that was made possible by speeding things up.
I'll see your senator, and I'll raise you two judges.
I have been shutting down more lately due to power costs and wanting to do my part to conserve energy. In any case, I am more interested in the laptop end of things. Virtually no one leaves a laptop up and running all the time.
:)
Did you actually stop to figure out the costs of running your computer 24/7? My friend gave me the same argument, and I calculated that the computer itself, to run 24/7, amounted to about $3USD/mo.
As for the laptop, no, but you have 2 easy options : suspend and hibernate. Almost all laptops support both, and have for many years now, and I always just put my laptop on suspend mode, and in the unlikely event that it will be off for more than a day, or away from a power outlet for more than a day, I'll just hibernate it.
If God gave us curiosity
last time I checked (yesterday actually) IE had a 86% market share. I'm running a Windows 2000 server. If your browsers don't implement the div tag with css style codes properly, ain't my fault =).
But in any case, I will attempt to correct the problem. Since it's just cosmetic, it's not very high on my priority list =(.
Thanks for the heads-up.
If God gave us curiosity
yessir it did. We pulled it down, and it rebooted back up just fine. The dept wouldnt fork out cash for a new product (To track a car fleet, of all things) so they just "turned the date back ten years". Yeah. That wont hurt, right?
morons.
Video meliora proboque deteriora sequor - Ovidius