Read the equation. Unless everyone is using a different program, your machine usage is going to be less than with an individual PC. Less hardware == Less $$$.
And I know this from experience. I used to admin a 40 user Citrix machine. It was a dual 233 PPro w/512 megs of RAM and a mirrored 40 GIG drive. Everyone ran MS Office, shipping apps, mainframe access software, web browsers, email, etc. without any performance problems. The only time they complained was in the (relatively rare) situation that we had to restart the machine due to some Windows instability.
Your average user does NOT make use of their machine's CPU for any other purpose than games. The real killer to their performance is memory. They wouldn't need so much memory if they didn't run their own copy of everything. Ergo, a shared environment means big savings for the users, big savings in admin costs, and more effective use of purchased hardware.
How is LTSP different from a diskless machine + a boot CD/Flash Drive/USB Drive with XDMCP query in the startup scripts?
Reading up on the LTSP "theory of operation", it seems that the above is EXACTLY how it works. I'm afraid that I don't understand what's new here. The addition of the MickeySoft "Terminal Services" terminology?
Now if only PCs could get Network Booting and hardware drivers into the firmware instead of having to tote around an extra data storage drive. Oh wait. PCs don't have firmware. $%(^!!! That's it! I'm going back to buying NEOStations.
As has been pointed out your running a "fat" OS on probably and even "fatter" machine. With a LTSP Server and a machine from disklessworkstations.com you just plug a station in a go. Your setup requires and actual OS to be hand installed and also requires a hard disk.
Let me ask this a different way then. How is LTSP different from a diskless machine + a boot CD/Flash Drive/USB Drive with XDMCP query in the startup scripts?
While many managers LOVED their ultra-fancy, green screen, terminals (despite never figuring out how to use all the extra features they paid for), they associate the idea with "old-tech". Managers want to be on the cutting edge. They perceive the cutting edge to be the tremendous waste of computing resources we call a PC. They don't have the background to understanding that a shared machine means the following memory usage pattern:
Let B = Size of binaries on disk Let D = Size of working data set Let U = Number of users for the given application
Memory = B + (D * U)
Instead they'd rather have:
Memory = (B + D) * U
Since the binary sizes are often significantly larger than the data sets (MB vs KB), the former simply makes more sense. But since it isn't as "cool", it doesn't catch on.
It must be my home-schooling background, but I've always looked for the information *before* I needed it. CompSci tends to be full of little threads you can pull on to open tremendous number of doors. For example, when I was growing up, 3D programming was the rage. (We didn't have no stinkin' "Graphics Accelerators" and we liked it that way! Oh, and we had to walk uphill through snow.) Just trying to learn the simple matter of drawing a polygon to the screen involved a massive number of algebraic and trigonometric concepts. By the time I hit Trig in high school, I was ready to learn because I was already using it!:-)
I suppose the real start I got was a college textbook on data structures that my Mother gave me when I was 11. Concepts like linked lists and bubble sorts may seem bleeding obvious, but forcing one's self to focus on the concept and calculate its best/worst case imparts a knowledge foundation that improves your work both consciously and unconsciously. By the time I got to Operating System Design, I understood that I shouldn't skip over the section on Round Robin Scheduling, because the author was demonstrating how new scheduling algorithms get derived.
Don't wait for the knowledge to hit you over the head. Go seek it out! Data Structures, OS Theory, Compiler Theory, Parser Theory, Parallel Computing Theory, Multidimensional Computations, Compression Theory, Encryption Theory, etc. are all in print and on the Internet, just waiting for you to learn them!
While you're at it, throw in Physics, History, and Literature. Physics is important because much of Computational Theory is based on the strict operation and limits of the Universe. History is important to understanding the role that your work will play in society and how things can be better improved. It can also keep you from making stupid mistakes. Literature is useful for expanding your mode of thinking. Don't think small. THINK BIG. Think big enough, and big problems will become smaller and smaller.
Oh, and don't be afraid to attack other areas of interest. You'd be amazed at how much linkage exists between the various scientific and engineering disciplines.:-)
It came down to me and one other fellow, and they went with the other fellow because he "had more experience"
Do you know what the other fellow's credentials were? It's quite possible that they hired him for reasons other than experience. I'd actually go as far as to say that in my experience it can be difficult to get a job with Accenture if you *are* well trained in your field.
When you get into the field (if you aren't already), ask a few experienced IT managers what they think about Accenture. They'll tend to tell you that Accenture never once delivered a working product, but instead pulled out so that the IT department could "finish the work and save money". After that, the project was invariably dismantled.
Accenture values the ability to make the client think that they're getting something for their money. The actual delivery is a mixed bag of whoever happens to be assigned on the project at a given time. Usually this means that you have one manager to every three programmers, and then managers to manage the managers. Each of these warm bodies comes in at $200+ per hour.
I had the distinct displeasure of being on the only successful Accenture project known to any of my superiors or colleagues. You see, my superiors had more experience with Accenture than their superiors. As a result, they knew what a mess the project was becoming. So they hired myself and a few other crack developers, and we TRAINED the Accenture coders. Most of the real work was still done (or redone as the case may be) by those of us who knew what we were doing. We then farmed out the tedious work to the ever-shrinking pool of Accenture employees.
We delivered on time, but the software ended up needing some major rework later on. Too many poor assumptions were made early in development that we weren't able to correct in time.
Of course, this is simply my experience. It's possible that Accenture is attempting to change itself to better meet the needs of the recession. I somewhat doubt it, however.
Hardware can scale almost linearly (well not quite but close).
Our ability to scale the speed of serial computations is primarily limited by the physical constraints of the universe. e.g. Light travels at a set speed. In order to make a computation happen faster, you have to make a more dense processor. As you start bumping up against the next limitation (e.g. computing at the molecular level) you have to get more clever about how you perform computations (e.g. Quantum computing.) Getting more clever requires application of new physics to computational sciences.
Scaled to infinity? Not likely. He would eventually ran into 'transmisions' problems where he did not have the bw to even feed the nodes or have them return the results somewhere meaningfull.
That was basically my point. He claimed he was using a super-powerful cluster to do exhaust modeling for a nuclear rocket. While he may have done such a thing (probably with incorrect assumptions in the physics model), he certainly did not do it on a cluster of the size he had claimed. Alternatively, he may have made up the performance numbers because he didn't know what the hell he was doing.
Programming for me is all about logic, and for some reason, the things that trip me up in actual math don't come into play there.
Remember Geometry? If you're anything like me, you probably struggled through it. But at the end of the day, I'm willing to bet that it increased your problem solving skills and ability to think logically a hundred fold or better.:-)
Anybody that has spent time reviewing code from a 'self taught coder' knows what I'm talking about (no offense to the self taught coders.)
Ooo... you were doing so well in your reply. Right up to the point that you insulted me.:-/
Sorry, I just couldn't resist the poke.:-) The "self-taught" coders that you think of are people who I think of as untrained junior programmers. For example, take just about anyone from Accenture (formerly Anderson Consulting). These are kids off the street who were looking for a job. Accenture says, "want to do computers?" and they go for it. Many of them can become halfway decent developers with the proper guidance and training, but unless you hire them off of Accenture you lose a lot of work once they leave.
This tends to get people in the mindset that they should only hire CompSci graduates. Unfortunately, that's also false economy. With many schools dumbing down their courses (not to mention the people who basically cheat their way through it to make $$$), degreed idiots are becoming very common in fields of science and engineering.
In the end, you simply have to apply critical thinking on any hiring candidates. Run them through a technical conversation and see how well they keep up. Degree or no degree, the ones who understand will be able to participate. The ones who don't understand will just agree with everything you say.
Am I the only one unable to find anything about a Lumix named linux distribution?
It is just a name for their custom Linux From Scratch compile. You won't find it because it isn't a real distro. However, you can visit the library to use it.:-)
What type of programming do you do, exactly, where the algorithm isn't the major focus of your work?
Not to defend the guy, but sometimes people use their knowledge unconsciously. For example, when I worked out a new collision detection algorithm for 2D games, I didn't need to write down the Big-O formula because it was straightforward enough for me to visualize the equation in my head. It took me a lot of time to realize that I did that,
One has to wonder how many CompSci majors have claimed that it was a waste of time, then went off to use their education in the back of their head.
Recently a "nuclear scientist" (the guy was a nut) was rather disparaging of Computer Sciences in an open forum. (Made the standard comments about "not a REAL science", blah, blah). He then went on to claim that in his younger days he had built the earliest computer cluster and that the software he wrote was so powerful that the cluster could have scaled up to infinity.
Suffice it to say, it felt really good to throw Amdahl's Law at him and call *bullshit*. He didn't talk to me after that.:-)
Let me second this. A good software engineer still needs to know the math behind what he's doing. Intuition can go a long ways, but sometimes intuition is wrong. And sometimes there just isn't a good way to develop something without invoking a few formulas. (Anyone who's developed a file system, say "Aye"!) The bright side is that most engineering work doesn't really require anything more complex than high school math. (Assuming that they're still *teaching* high school math. Grrr...)
My best suggestion would be to shy away from any school that skips over the theory and math behind computational sciences. I myself ended up skipping the degree all together and picked up the various textbooks and papers necessary to educate myself in the field. That's not the path for everyone (especially if you're not very self-motivated), but for me it was better than trying to sort through which school was teaching the real thing and which one was dumbing it down to improve attendance.
...how much my software has changed. It used to be that I'd use MS Office, Internet Explorer, Paintshop Pro, WinZIP, and other for-pay applications to get my work done. These days I find myself using FireFox, OpenOffice, JEdit, NetBeans, Cygwin, EnZIP, GIMP, and other Open Source tools. And nearly every one of them is superior to the application I replaced. Fascinating world we live in.
1. Sun bundles stuff that other distros can't. e.g. StarOffice (which has features that OpenOffice doesn't), Macromedia Flash, Java, RealPlayer, etc.
2. $100 is the base price (actually $50 right now) that gets you one year of updates. You never have to pay another dime unless you want ANOTHER year of updates.
3. Sun's plan is to bundle all of their desktop software into one package. As they add new value (e.g. MSAccess support for StarOffice is in the works), you get those upgrades free.
4. Unlike many ad-hoc distros, JDS is founded upon the idea of being a consolidated desktop.
I will say that at this point JDS is only slightly *different* from SuSE. SuSE may actually be the better pick. However, I do expect that JDS will become a lot more competitive in the future.
In writing dates, though, it's definitely dd/mm/yy.
Screw that. I'd much rather have yyyy/mm/dd. It's easier to read (since you know that you're in the right year before you get to the month) and it sorts much easer.
The NTFS module is not included at all. RedHat considers it to be "non-free" software and thus refuses to ship it. There is an RPM and SRPM for Fedora Core 1 from the Linux-NTFS project, but FC2 is completely unsupported.
It sounds like they're trying to undercut Sun's Java Desktop System. Sun is selling a bundle of their software (JDS/StarOffice/Java Media Player/Java Development tools/etc.) for $100 per desktop per year. The $100 is for maintenance and upgrades, so you can stop paying at any time and keep using your current desktop.
The CDRWs were used in the same drive that they were burned. There were no software errors related to corrupted data. In fact, most installs were done over the Internet, so it's a bit hard to blame the CDRW disks for anything.
Slashdot Mirror
I love the number of shares requested! "31337", Priceless!
Read the equation. Unless everyone is using a different program, your machine usage is going to be less than with an individual PC. Less hardware == Less $$$.
And I know this from experience. I used to admin a 40 user Citrix machine. It was a dual 233 PPro w/512 megs of RAM and a mirrored 40 GIG drive. Everyone ran MS Office, shipping apps, mainframe access software, web browsers, email, etc. without any performance problems. The only time they complained was in the (relatively rare) situation that we had to restart the machine due to some Windows instability.
Your average user does NOT make use of their machine's CPU for any other purpose than games. The real killer to their performance is memory. They wouldn't need so much memory if they didn't run their own copy of everything. Ergo, a shared environment means big savings for the users, big savings in admin costs, and more effective use of purchased hardware.
How is LTSP different from a diskless machine + a boot CD/Flash Drive/USB Drive with XDMCP query in the startup scripts?
Reading up on the LTSP "theory of operation", it seems that the above is EXACTLY how it works. I'm afraid that I don't understand what's new here. The addition of the MickeySoft "Terminal Services" terminology?
Now if only PCs could get Network Booting and hardware drivers into the firmware instead of having to tote around an extra data storage drive. Oh wait. PCs don't have firmware. $%(^!!! That's it! I'm going back to buying NEOStations.
As has been pointed out your running a "fat" OS on probably and even "fatter" machine. With a LTSP Server and a machine from disklessworkstations.com you just plug a station in a go. Your setup requires and actual OS to be hand installed and also requires a hard disk.
Let me ask this a different way then. How is LTSP different from a diskless machine + a boot CD/Flash Drive/USB Drive with XDMCP query in the startup scripts?
I loved Geometry! And Linear Algebra!
:-)
Some Math hater you turned out to be! Congratulations, you've been using mathematics and didn't even know it!
It must be my home-schooling background, but I've always looked for the information *before* I needed it. CompSci tends to be full of little threads you can pull on to open tremendous number of doors. For example, when I was growing up, 3D programming was the rage. (We didn't have no stinkin' "Graphics Accelerators" and we liked it that way! Oh, and we had to walk uphill through snow.) Just trying to learn the simple matter of drawing a polygon to the screen involved a massive number of algebraic and trigonometric concepts. By the time I hit Trig in high school, I was ready to learn because I was already using it! :-)
:-)
I suppose the real start I got was a college textbook on data structures that my Mother gave me when I was 11. Concepts like linked lists and bubble sorts may seem bleeding obvious, but forcing one's self to focus on the concept and calculate its best/worst case imparts a knowledge foundation that improves your work both consciously and unconsciously. By the time I got to Operating System Design, I understood that I shouldn't skip over the section on Round Robin Scheduling, because the author was demonstrating how new scheduling algorithms get derived.
Don't wait for the knowledge to hit you over the head. Go seek it out! Data Structures, OS Theory, Compiler Theory, Parser Theory, Parallel Computing Theory, Multidimensional Computations, Compression Theory, Encryption Theory, etc. are all in print and on the Internet, just waiting for you to learn them!
While you're at it, throw in Physics, History, and Literature. Physics is important because much of Computational Theory is based on the strict operation and limits of the Universe. History is important to understanding the role that your work will play in society and how things can be better improved. It can also keep you from making stupid mistakes. Literature is useful for expanding your mode of thinking. Don't think small. THINK BIG. Think big enough, and big problems will become smaller and smaller.
Oh, and don't be afraid to attack other areas of interest. You'd be amazed at how much linkage exists between the various scientific and engineering disciplines.
It came down to me and one other fellow, and they went with the other fellow because he "had more experience"
Do you know what the other fellow's credentials were? It's quite possible that they hired him for reasons other than experience. I'd actually go as far as to say that in my experience it can be difficult to get a job with Accenture if you *are* well trained in your field.
When you get into the field (if you aren't already), ask a few experienced IT managers what they think about Accenture. They'll tend to tell you that Accenture never once delivered a working product, but instead pulled out so that the IT department could "finish the work and save money". After that, the project was invariably dismantled.
Accenture values the ability to make the client think that they're getting something for their money. The actual delivery is a mixed bag of whoever happens to be assigned on the project at a given time. Usually this means that you have one manager to every three programmers, and then managers to manage the managers. Each of these warm bodies comes in at $200+ per hour.
I had the distinct displeasure of being on the only successful Accenture project known to any of my superiors or colleagues. You see, my superiors had more experience with Accenture than their superiors. As a result, they knew what a mess the project was becoming. So they hired myself and a few other crack developers, and we TRAINED the Accenture coders. Most of the real work was still done (or redone as the case may be) by those of us who knew what we were doing. We then farmed out the tedious work to the ever-shrinking pool of Accenture employees.
We delivered on time, but the software ended up needing some major rework later on. Too many poor assumptions were made early in development that we weren't able to correct in time.
Of course, this is simply my experience. It's possible that Accenture is attempting to change itself to better meet the needs of the recession. I somewhat doubt it, however.
Hardware can scale almost linearly (well not quite but close).
Our ability to scale the speed of serial computations is primarily limited by the physical constraints of the universe. e.g. Light travels at a set speed. In order to make a computation happen faster, you have to make a more dense processor. As you start bumping up against the next limitation (e.g. computing at the molecular level) you have to get more clever about how you perform computations (e.g. Quantum computing.) Getting more clever requires application of new physics to computational sciences.
Scaled to infinity? Not likely. He would eventually ran into 'transmisions' problems where he did not have the bw to even feed the nodes or have them return the results somewhere meaningfull.
That was basically my point. He claimed he was using a super-powerful cluster to do exhaust modeling for a nuclear rocket. While he may have done such a thing (probably with incorrect assumptions in the physics model), he certainly did not do it on a cluster of the size he had claimed. Alternatively, he may have made up the performance numbers because he didn't know what the hell he was doing.
Programming for me is all about logic, and for some reason, the things that trip me up in actual math don't come into play there.
:-)
Remember Geometry? If you're anything like me, you probably struggled through it. But at the end of the day, I'm willing to bet that it increased your problem solving skills and ability to think logically a hundred fold or better.
Anybody that has spent time reviewing code from a 'self taught coder' knows what I'm talking about (no offense to the self taught coders.)
:-/
:-) The "self-taught" coders that you think of are people who I think of as untrained junior programmers. For example, take just about anyone from Accenture (formerly Anderson Consulting). These are kids off the street who were looking for a job. Accenture says, "want to do computers?" and they go for it. Many of them can become halfway decent developers with the proper guidance and training, but unless you hire them off of Accenture you lose a lot of work once they leave.
Ooo... you were doing so well in your reply. Right up to the point that you insulted me.
Sorry, I just couldn't resist the poke.
This tends to get people in the mindset that they should only hire CompSci graduates. Unfortunately, that's also false economy. With many schools dumbing down their courses (not to mention the people who basically cheat their way through it to make $$$), degreed idiots are becoming very common in fields of science and engineering.
In the end, you simply have to apply critical thinking on any hiring candidates. Run them through a technical conversation and see how well they keep up. Degree or no degree, the ones who understand will be able to participate. The ones who don't understand will just agree with everything you say.
Am I the only one unable to find anything about a Lumix named linux distribution?
:-)
It is just a name for their custom Linux From Scratch compile. You won't find it because it isn't a real distro. However, you can visit the library to use it.
What type of programming do you do, exactly, where the algorithm isn't the major focus of your work?
Not to defend the guy, but sometimes people use their knowledge unconsciously. For example, when I worked out a new collision detection algorithm for 2D games, I didn't need to write down the Big-O formula because it was straightforward enough for me to visualize the equation in my head. It took me a lot of time to realize that I did that,
One has to wonder how many CompSci majors have claimed that it was a waste of time, then went off to use their education in the back of their head.
Recently a "nuclear scientist" (the guy was a nut) was rather disparaging of Computer Sciences in an open forum. (Made the standard comments about "not a REAL science", blah, blah). He then went on to claim that in his younger days he had built the earliest computer cluster and that the software he wrote was so powerful that the cluster could have scaled up to infinity.
:-)
Suffice it to say, it felt really good to throw Amdahl's Law at him and call *bullshit*. He didn't talk to me after that.
Let me second this. A good software engineer still needs to know the math behind what he's doing. Intuition can go a long ways, but sometimes intuition is wrong. And sometimes there just isn't a good way to develop something without invoking a few formulas. (Anyone who's developed a file system, say "Aye"!) The bright side is that most engineering work doesn't really require anything more complex than high school math. (Assuming that they're still *teaching* high school math. Grrr...)
My best suggestion would be to shy away from any school that skips over the theory and math behind computational sciences. I myself ended up skipping the degree all together and picked up the various textbooks and papers necessary to educate myself in the field. That's not the path for everyone (especially if you're not very self-motivated), but for me it was better than trying to sort through which school was teaching the real thing and which one was dumbing it down to improve attendance.
...how much my software has changed. It used to be that I'd use MS Office, Internet Explorer, Paintshop Pro, WinZIP, and other for-pay applications to get my work done. These days I find myself using FireFox, OpenOffice, JEdit, NetBeans, Cygwin, EnZIP, GIMP, and other Open Source tools. And nearly every one of them is superior to the application I replaced. Fascinating world we live in.
*sigh* Please read the comments on the reviews. I've already addressed the FreeBSD vs. Debian vs. Gentoo vs. Linux for my Dog.
But the Java Desktop requires Redhat Linux 7.3?
No, the Desktop is SuSE based. The admin interface for the Desktop cluster requires RedHat 7.3. They were a bit unclear on that.
1. Sun bundles stuff that other distros can't. e.g. StarOffice (which has features that OpenOffice doesn't), Macromedia Flash, Java, RealPlayer, etc.
2. $100 is the base price (actually $50 right now) that gets you one year of updates. You never have to pay another dime unless you want ANOTHER year of updates.
3. Sun's plan is to bundle all of their desktop software into one package. As they add new value (e.g. MSAccess support for StarOffice is in the works), you get those upgrades free.
4. Unlike many ad-hoc distros, JDS is founded upon the idea of being a consolidated desktop.
I will say that at this point JDS is only slightly *different* from SuSE. SuSE may actually be the better pick. However, I do expect that JDS will become a lot more competitive in the future.
In writing dates, though, it's definitely dd/mm/yy.
Screw that. I'd much rather have yyyy/mm/dd. It's easier to read (since you know that you're in the right year before you get to the month) and it sorts much easer.
The NTFS module is not included at all. RedHat considers it to be "non-free" software and thus refuses to ship it. There is an RPM and SRPM for Fedora Core 1 from the Linux-NTFS project, but FC2 is completely unsupported.
[RedHat Desktop is] $5/month
It sounds like they're trying to undercut Sun's Java Desktop System. Sun is selling a bundle of their software (JDS/StarOffice/Java Media Player/Java Development tools/etc.) for $100 per desktop per year. The $100 is for maintenance and upgrades, so you can stop paying at any time and keep using your current desktop.
The CDRWs were used in the same drive that they were burned. There were no software errors related to corrupted data. In fact, most installs were done over the Internet, so it's a bit hard to blame the CDRW disks for anything.
Apparently, no one paid attention to the review:
1. I was using Fedora Core 2. There is NO NTFS support for Fedore Core 2. Fedora Core 1 was only recently added.
2. The whole point of the review was that it would work out of the box. Fedora didn't, so I dumped it. Boo hoo.
It just so happens that SuSE 9 was the only distro that met my needs. Seems that SuSE has the jump on everyone else. :-)