Consider Moore's Law for a minute, in the context of power consumption (from your post.)
Your current desktop is about 1,000x more powerful than the desktops we had about 20 years ago. It has 8,000x more memory and 50,000x more hard drive space. And yet it is still plugged into the same wall plug it was back then. Same amount of energy, plus or minus (discounting a crazy overclocked CPU / video card - but still you can get away with plugging it into the same wall socket.)
I don't have a point, just considering something out loud.
You know, someone who's actually been in the IT industry for ten, twenty years knows that in ten, twenty years you can buy the capacity in their datacenter for $100 and keep it in your pocket.
Well, maybe not that bad, but you get the drift.
Actually yes, that bad. In comparing my home system to my very first home system, bought roughly 17 years ago, I find that I have roughly 1,000x more CPU (based purely on clock speed, not taking into account pipelining, caching, etc), 8,000x more memory, and 50,000x more hard drive space.
If things keep progressing at that speed, and there's no reason to think they won't (single chips won't, but parallelizing is already taking up the slack) and in 20 years a high end home system will have... well the CPU power of 1,000 current machines, the memory currently spread across 8,000 machines, and the drive space currently spread across 50,000 machines. In other words, in twenty years that entire data center will have less horsepower and capacity than some of the systems we will have already retired and are probably using to hold open doors. You will offer to give away machines with more capacity / horsepower than that entire data center, and nobody will take it away for free.
Where do I think systems are going? Unified memory model using massive quantities of persistent memory that does not need refreshing and is as fast as current 'core' memory. Once flash memory is replaced with something quite a bit faster, the line between operational memory and long term storage blurs - and the line becomes 'working storage' vs 'committed updates'. Throw in an inference engine that works across the entire data stream, and give it the ability to dynamically determine what it is trying to determine based on the data in the stream - and BINGO : that's where we are headed.
I think he was talking about performance, not syntax. I had a customer once spec that the system needed to sustain 7 transactions a second on a real-time system. Sure thing. No Problem. They said they wanted it in Java, at which point I laughed. This was in late 2000 or early 2001. I suggested we use C, and they bought it.
And there is that minor thing about being able to write operating system extensions in C/C++ and Assembly (but not in Java), but I doubt that was what he was thinking.
Hah. Yea that's 100% true, but think about all of us that already have an education and skills - we're going to be RICH! Somehow I don't think I've thought my cunning plan all the way through, but it sure looks good on paper.
Actually the money is taken at gun-point from the rich / middle class and loaned to the poor, who then get college degrees with which they get real jobs and become middle class, pay that loan back with interest, and pay massive taxes on their own thus reducing the tax burden on the original rich/middle class.
I originally got about $9,000 in student loans getting my undergrad degree. Paid ALL of it back, at about 6% interest, over the course of about 8 or 9 years. Worked my way up to a decent job, and I have paid $12k~$15k in income taxes per year every year for the past 5, with lesser amounts before then. I figure the $9,000 money 'given' to a poor kid to go to college has paid back in spades - it was paid back 106% the first 8 years and has paid back over 100% per year since then. If all tax revenues were invested in ways that paid back 100% after 8 years and 100% PER YEAR thereafter, forever - well do the math.
Funny - I'm writing this from within openSuSE 10.3 running (you guessed it) in a VMWare session on a WinXP box. Works incredibly well, welcome to the club. My DHCP is generally pretty quick, but I'm another one of those guys plugged directly into a Linksys firewall/router, which also assigns DHCP addresses inside the firewall. Less than 10 seconds, generally, to get an IP - but as someone else pointed out I could probably hardcode the IP address and skip that step altogether for boot performance.
That said, I leave the SuSE desktop running 24x7 so the boot time isn't that much of an issue.
Actually most of the 30-90 second boot process is interrogating the hardware from top to bottom to re-discover all the different pieces of hardware you might have in the system, on a one-by-one basis, then loading the individual sets of drivers for each. Just because you had an nVidia video card, Intel chipset, Creative Labs PCI sound card, six USB ports, and a 100Mb/s Ethernet NIC by SMC in your system when you shut it down last time, that doesn't mean those are the components in your system this time you do a cold boot - so the OS goes through the entire discovery process from top to bottom again, loading the drivers for the hardware it finds after going through the process of discovering each piece of hardware in your system.
That's why my laptop can recovery from 'stand-by' or 'hibernate' modes in almost no-time, but a cold boot still takes a veritable lifetime - approaching two minutes before the system is fully loaded and operational, and why solid state drives only shave 6-8 seconds from boot times while offering nano-second seek times.
Back in the old days you hard-coded the memory addresses, IRQs, DMA addresses, etc of the hardware in the boot files, which is why older (much slower) systems booted so fast. No parsing every IRQ and memory address in the system looking for new hardware, asking each piece of hardware 'what are you, what kind of drivers do you use, which IRQ do you want... no, that one is used, how about picking a different one...'. If there was a way to code Windows to skip the PnP (Plug and Play) and just tell it what all you had for hardware and where in memory / IRQ / DMAland to put each piece (or just tell it that the hardware it had when it shut down was the same hardware it had the next boot) - I'm guessing that the OS would boot a LOT faster. Like order of magnitude faster.
Actually, if memory serves - in the beginning of the movie, with his girlfriend watching, he says something about 'I didn't deserve that grade' (or something to that effect) and changes one of his grades.
Tell you what - you can start believing that the government is acting in your (professional Americans working in IT) best interests the day AFTER they shut down the H1-B program. 1.5 million foreigners were given jobs over eight years under Clinton, that's 1.5M jobs that weren't filled by Americans. How many software engineering positions are there in total in the USA? Getting a clue about where we are on the food chain yet?
I don't know if / how it will get better, but hoping the government is going to step up and make things better for IT workers in the US is a 2400 baud pipe dream.
So what you are saying is - if you rely solely on Windows, you are probably going to crash on a regular basis, and more frequently on some hardware than others?
Newsflash - everyone does have a price. If one person doesn't have a price, there are reams of people that do have a price, a will kill that first guy and his entire family for that price. The first guy will obviously sell whatever it is he says 'isn't for sale at any price' in order to keep himself and his entire family alive.
Ask that Russian spammer guy. His price was pretty high, but there was one group willing to pay it.
Actually the System S, as described to me, was designed to take non-linear and dimensionless (rather, undefined raw data without predefined dimensions) and do heuristic analysis looking for trends. The trick would be to describe the technical rules and the potential end game results, and let the System run for a while playing against itself in order to come up with the basic analysis, then let it watch a bunch of games played by humans and try to predict the results as the game plays.
Just curious - a Blue Gene Series L, Series P, or one of the new Series Q's upstairs? If you've got a Q, are they hiring?
I'd like to feed this problem to one of the new IBM System S machines and give it a crack. This is exactly the kind of problem-set the Sys/S boxes were designed for...
More than a year of uptime? That's child's play for Netware. I've seen plenty of boxes with several years uptime - all running Netware 4.x I've heard stories of Netware servers that got lost, physically misplaced (one, according to legend, was drywalled into a building by a work-crew that didn't know it was there) that ran for a very long time (years) without anybody knowing where they were.
It pretty much takes a hardware failure to bring down a Netware box. A bad cpu fan killed one of mine, and a bad power supply killed another (both in the late '90s).
BTW - thanks for sharing the last access update regedit hack. Got any other 30% performance gain tips you want to share, because I will take as many of those as you're willing to share. If I hadn't just replied, I'd be ponying up a mod point or two.
Having spent hours, days, years studying the effects of hard drive defragmentation, let me put the kibosh on 'intelligent defragmentation' here and now. Defragmenting the files themselves gives about 20% of the potential benefits of defragmentation. Defragmenting the file allocation table (FAT on FAT/FAT32 file systems, or MFT on NTFS file systems) gives the remaining 80% of the performance boost potentially given by defragging.
In the big scheme of things, it honestly doesn't matter whether the most recently used files are at the beginning of the drive, next to each other, or on opposite sides of the drive - if the file allocation table (or MFT) is sufficiently fragmented. Frag out the FAT/MFT bad enough over time, and simply defragging the MFT/FAT will make your computer run an order of magnitude faster.
Want the bad news? Windows doesn't ship with a FAT/MFT defragger (well through XP. Not sure about Vista.) Only way I know to do it is with aftermarket software like Diskeeper (excellent product, BTW, 99% of the time.)
Crap, I was thinking scotch tape on the NIC card on the inside of the machine, meaning inside the contacts that go inside the ISA or PCI card slot. If he is talking about on the outside (on the RJ-45, and although I've never heard of a RJ-15, I have diagnosed plenty of problems with RJ-11 jacks) - heck yea, those jacks corrode all the time from air exposure - that's something I would have caught (the LEDs on the back of the NIC never would have lit up, and the jack would have been investigated in a pretty good hurry.
Sounds like cello-tape, used on the inside of the card-slot like I was thinking, would have thrown you off too.
Regarding the tape on the NIC card contacts - there's the rub, odds are I would never have caught it either. It's not a normally occuring fault that a 'real world' tech would ever experience in his lifetime. Any NIC that has been in use (successfully) long enough to gather dust has survived the infant mortality period and because they plug into only one other device (switch, generally) - if the device on the other end doesn't explode due to a crazy electrical storm, it isn't going to send a bad enough spike to smoke the NIC. IMHO, once a network card has been proven good, it never dies. Well almost never, assuming high quality NICs in the first place - I've installed somewhere in the neighborhood of 400 SMC/3COM network cards and have never had a failure once a NIC passed the first few weeks successfully.
Of course if it wasn't talking to the network and wasn't showing up in the device manager, I would have popped out the NIC to replace it (thinking it was bad) and when I saw the scotch tape on it, I probably would have whacked someone over the head with it.
Actually space travel boils down to 3D vector math, and in order to change directions the ship must both accelerate in the direction it wants to go, and decellerate in the direction it was once going. Thus, yes it would need to slow down in the direction it was once going, lest it continue to travel in that direction (in addition to the direction it turns towards by accelerating in that direction.) The pilot could do this before, during, or after the turn, but unless the travel in the original direction is offset, the craft will maintain the speed in the direction it was originally travelling in (quite possibly in addition to the new speed in the new direction.)
"In his interview, Burgett points out that the people working for him are also recovering drug addicts or recovering mental illness patients" Or as we call them on Slashdot, 'Linux System Administrators'.
Slashdot Mirror
Consider Moore's Law for a minute, in the context of power consumption (from your post.)
Your current desktop is about 1,000x more powerful than the desktops we had about 20 years ago. It has 8,000x more memory and 50,000x more hard drive space. And yet it is still plugged into the same wall plug it was back then. Same amount of energy, plus or minus (discounting a crazy overclocked CPU / video card - but still you can get away with plugging it into the same wall socket.)
I don't have a point, just considering something out loud.
You know, someone who's actually been in the IT industry for ten, twenty years knows that in ten, twenty years you can buy the capacity in their datacenter for $100 and keep it in your pocket.
... well the CPU power of 1,000 current machines, the memory currently spread across 8,000 machines, and the drive space currently spread across 50,000 machines. In other words, in twenty years that entire data center will have less horsepower and capacity than some of the systems we will have already retired and are probably using to hold open doors. You will offer to give away machines with more capacity / horsepower than that entire data center, and nobody will take it away for free.
Well, maybe not that bad, but you get the drift.
Actually yes, that bad. In comparing my home system to my very first home system, bought roughly 17 years ago, I find that I have roughly 1,000x more CPU (based purely on clock speed, not taking into account pipelining, caching, etc), 8,000x more memory, and 50,000x more hard drive space.
If things keep progressing at that speed, and there's no reason to think they won't (single chips won't, but parallelizing is already taking up the slack) and in 20 years a high end home system will have
Where do I think systems are going? Unified memory model using massive quantities of persistent memory that does not need refreshing and is as fast as current 'core' memory. Once flash memory is replaced with something quite a bit faster, the line between operational memory and long term storage blurs - and the line becomes 'working storage' vs 'committed updates'. Throw in an inference engine that works across the entire data stream, and give it the ability to dynamically determine what it is trying to determine based on the data in the stream - and BINGO : that's where we are headed.
I think he was talking about performance, not syntax.
I had a customer once spec that the system needed to sustain 7 transactions a second on a real-time system. Sure thing. No Problem.
They said they wanted it in Java, at which point I laughed. This was in late 2000 or early 2001. I suggested we use C, and they bought it.
And there is that minor thing about being able to write operating system extensions in C/C++ and Assembly (but not in Java), but I doubt that was what he was thinking.
Hah. Yea that's 100% true, but think about all of us that already have an education and skills - we're going to be RICH!
Somehow I don't think I've thought my cunning plan all the way through, but it sure looks good on paper.
Actually the money is taken at gun-point from the rich / middle class and loaned to the poor, who then get college degrees with which they get real jobs and become middle class, pay that loan back with interest, and pay massive taxes on their own thus reducing the tax burden on the original rich/middle class.
I originally got about $9,000 in student loans getting my undergrad degree. Paid ALL of it back, at about 6% interest, over the course of about 8 or 9 years. Worked my way up to a decent job, and I have paid $12k~$15k in income taxes per year every year for the past 5, with lesser amounts before then. I figure the $9,000 money 'given' to a poor kid to go to college has paid back in spades - it was paid back 106% the first 8 years and has paid back over 100% per year since then. If all tax revenues were invested in ways that paid back 100% after 8 years and 100% PER YEAR thereafter, forever - well do the math.
Funny - I'm writing this from within openSuSE 10.3 running (you guessed it) in a VMWare session on a WinXP box. Works incredibly well, welcome to the club.
My DHCP is generally pretty quick, but I'm another one of those guys plugged directly into a Linksys firewall/router, which also assigns DHCP addresses inside the firewall. Less than 10 seconds, generally, to get an IP - but as someone else pointed out I could probably hardcode the IP address and skip that step altogether for boot performance.
That said, I leave the SuSE desktop running 24x7 so the boot time isn't that much of an issue.
Actually most of the 30-90 second boot process is interrogating the hardware from top to bottom to re-discover all the different pieces of hardware you might have in the system, on a one-by-one basis, then loading the individual sets of drivers for each. Just because you had an nVidia video card, Intel chipset, Creative Labs PCI sound card, six USB ports, and a 100Mb/s Ethernet NIC by SMC in your system when you shut it down last time, that doesn't mean those are the components in your system this time you do a cold boot - so the OS goes through the entire discovery process from top to bottom again, loading the drivers for the hardware it finds after going through the process of discovering each piece of hardware in your system.
... no, that one is used, how about picking a different one...'. If there was a way to code Windows to skip the PnP (Plug and Play) and just tell it what all you had for hardware and where in memory / IRQ / DMAland to put each piece (or just tell it that the hardware it had when it shut down was the same hardware it had the next boot) - I'm guessing that the OS would boot a LOT faster. Like order of magnitude faster.
That's why my laptop can recovery from 'stand-by' or 'hibernate' modes in almost no-time, but a cold boot still takes a veritable lifetime - approaching two minutes before the system is fully loaded and operational, and why solid state drives only shave 6-8 seconds from boot times while offering nano-second seek times.
Back in the old days you hard-coded the memory addresses, IRQs, DMA addresses, etc of the hardware in the boot files, which is why older (much slower) systems booted so fast. No parsing every IRQ and memory address in the system looking for new hardware, asking each piece of hardware 'what are you, what kind of drivers do you use, which IRQ do you want
Actually, if memory serves - in the beginning of the movie, with his girlfriend watching, he says something about 'I didn't deserve that grade' (or something to that effect) and changes one of his grades.
As long as Google survives, we're all set. There isn't much I ~can't~ do after about 15 minutes w/ Google.
Yea, good luck with that.
Tell you what - you can start believing that the government is acting in your (professional Americans working in IT) best interests the day AFTER they shut down the H1-B program. 1.5 million foreigners were given jobs over eight years under Clinton, that's 1.5M jobs that weren't filled by Americans. How many software engineering positions are there in total in the USA? Getting a clue about where we are on the food chain yet?
I don't know if / how it will get better, but hoping the government is going to step up and make things better for IT workers in the US is a 2400 baud pipe dream.
So what you are saying is - if you rely solely on Windows, you are probably going to crash on a regular basis, and more frequently on some hardware than others?
Newsflash - everyone does have a price.
If one person doesn't have a price, there are reams of people that do have a price, a will kill that first guy and his entire family for that price.
The first guy will obviously sell whatever it is he says 'isn't for sale at any price' in order to keep himself and his entire family alive.
Ask that Russian spammer guy. His price was pretty high, but there was one group willing to pay it.
Evidently murder isn't funny. Except when you murder a clown or a mime (then it is pretty funny.) Or a spammer, evidently.
Actually the System S, as described to me, was designed to take non-linear and dimensionless (rather, undefined raw data without predefined dimensions) and do heuristic analysis looking for trends. The trick would be to describe the technical rules and the potential end game results, and let the System run for a while playing against itself in order to come up with the basic analysis, then let it watch a bunch of games played by humans and try to predict the results as the game plays.
Just curious - a Blue Gene Series L, Series P, or one of the new Series Q's upstairs?
If you've got a Q, are they hiring?
Envision JumpMan, from a first person perspective using a recent engine.
Damn, that would be GOOD!
I'd like to feed this problem to one of the new IBM System S machines and give it a crack. This is exactly the kind of problem-set the Sys/S boxes were designed for...
More than a year of uptime? That's child's play for Netware. I've seen plenty of boxes with several years uptime - all running Netware 4.x
I've heard stories of Netware servers that got lost, physically misplaced (one, according to legend, was drywalled into a building by a work-crew that didn't know it was there) that ran for a very long time (years) without anybody knowing where they were.
It pretty much takes a hardware failure to bring down a Netware box. A bad cpu fan killed one of mine, and a bad power supply killed another (both in the late '90s).
A few hundred Gigs of system memory ought to be enough for anybody.
BTW - thanks for sharing the last access update regedit hack. Got any other 30% performance gain tips you want to share, because I will take as many of those as you're willing to share. If I hadn't just replied, I'd be ponying up a mod point or two.
Having spent hours, days, years studying the effects of hard drive defragmentation, let me put the kibosh on 'intelligent defragmentation' here and now.
Defragmenting the files themselves gives about 20% of the potential benefits of defragmentation.
Defragmenting the file allocation table (FAT on FAT/FAT32 file systems, or MFT on NTFS file systems) gives the remaining 80% of the performance boost potentially given by defragging.
In the big scheme of things, it honestly doesn't matter whether the most recently used files are at the beginning of the drive, next to each other, or on opposite sides of the drive - if the file allocation table (or MFT) is sufficiently fragmented. Frag out the FAT/MFT bad enough over time, and simply defragging the MFT/FAT will make your computer run an order of magnitude faster.
Want the bad news? Windows doesn't ship with a FAT/MFT defragger (well through XP. Not sure about Vista.)
Only way I know to do it is with aftermarket software like Diskeeper (excellent product, BTW, 99% of the time.)
Crap, I was thinking scotch tape on the NIC card on the inside of the machine, meaning inside the contacts that go inside the ISA or PCI card slot. If he is talking about on the outside (on the RJ-45, and although I've never heard of a RJ-15, I have diagnosed plenty of problems with RJ-11 jacks) - heck yea, those jacks corrode all the time from air exposure - that's something I would have caught (the LEDs on the back of the NIC never would have lit up, and the jack would have been investigated in a pretty good hurry.
Sounds like cello-tape, used on the inside of the card-slot like I was thinking, would have thrown you off too.
Regarding the tape on the NIC card contacts - there's the rub, odds are I would never have caught it either. It's not a normally occuring fault that a 'real world' tech would ever experience in his lifetime. Any NIC that has been in use (successfully) long enough to gather dust has survived the infant mortality period and because they plug into only one other device (switch, generally) - if the device on the other end doesn't explode due to a crazy electrical storm, it isn't going to send a bad enough spike to smoke the NIC. IMHO, once a network card has been proven good, it never dies. Well almost never, assuming high quality NICs in the first place - I've installed somewhere in the neighborhood of 400 SMC/3COM network cards and have never had a failure once a NIC passed the first few weeks successfully.
Of course if it wasn't talking to the network and wasn't showing up in the device manager, I would have popped out the NIC to replace it (thinking it was bad) and when I saw the scotch tape on it, I probably would have whacked someone over the head with it.
Actually space travel boils down to 3D vector math, and in order to change directions the ship must both accelerate in the direction it wants to go, and decellerate in the direction it was once going. Thus, yes it would need to slow down in the direction it was once going, lest it continue to travel in that direction (in addition to the direction it turns towards by accelerating in that direction.) The pilot could do this before, during, or after the turn, but unless the travel in the original direction is offset, the craft will maintain the speed in the direction it was originally travelling in (quite possibly in addition to the new speed in the new direction.)
"In his interview, Burgett points out that the people working for him are also recovering drug addicts or recovering mental illness patients"
Or as we call them on Slashdot, 'Linux System Administrators'.
Perhaps I am doing something wrong? ... Windows Ultimate <-- This.
My main windows box is