The current and past trend in databases is to make everything work in a SQL database. As far as commercial relational databases go SQL is the standard. Check out M. Stonebraker, "Inclusion of New Types in Relational Data Base Systems", IEEE Data Engineering Conference, Los Angeles, CA, February 1986. (http://citeseer.nj.nec.com/ncontextsummary/11868/ 52519). This is a paper on how to extend relational databases to handle all sorts of data. It has been used to create Object Relational databases, add in spatial information to databases, and many other types of data. The latests and greatest data type that people are trying to integrate is XML. This has not met with so much sucess when efficently querying the data, but there is a great deal of current research on the subject.
After spending literly 30 seconds reading the man page, and a day getting used to it, Ion was the best window manager I have ever used. It was designed to be very efficent, and worked great as a programming interface.
The point would be to have a small cluster (5 nodes would do 4 cluster + a login box) so you could learn how to code/admin that type of thing. All of the software is free and the hardware is free to you. So take a spin at learning MPI, OpenMOSIX, or sockets programming!
Currently the top 5 consist of 4 machines that have a Therotical maximum speed (Rmax) the are larger then the 10TFLOPS this machine will have. Then you have to translate that into peak speed which is what matter and what this list uses to rank the machines. Peak will be a good deal less, but this mostly has to do with the way the systems are interconnected and not the machines themselves. Say what you may about the G5 but the interconnect is more important.
There is only one machine in the top 5 that this cluster could beat. The rest of the world has had 6 months to build machines too.
This should be a top 10 machine for sure. Good to see more fast machines being built every day.
Bull Shit. Doing science (be it playing with chemicals or computers) just for the sake of doing science is what has given us MANY of our most useful tools. Penicillin and Lasers were discovered because some scientist was messing around with something and said "Wait, why is that happening?". We also get things that are just part of the journey to destination. This would include things like Velcro.
Besides that what is so wrong about people trying to do things cheaper? More reliable? Because they are interested in them?
If they can make this go as fast as (or close) the theory says, this would be a pretty good platform. Keep in mind that we will be seeing clustering of different types of systems for a long time to come. The more research we do the better we will be at it. In the end it will allow us to be more efficent.
I started out with a copy of truespace a a book. I worked with that for a couple of months. My first job was doing some 3-D modeling and animation for a presentation. I made a little bit of money then "upgraded" to 3DSMAX.
When I say "upgraded" it was an upgrade but nothing that was beyond belief. Yes it was multi-threaded and could do network rendering, there were some other things that it could handle, but it took away productuvity for some time while I learned the new system.
What I would suggest to you is to pick a system that you can afford (or "borrow") and get yourself a book or two. Have some ideas of what you would like to do for a project and learn how to get there.
It takes a lot of time and, for me, a change in how I interacted with the computer to become good enough to claim my work as professional.
One of the other things you might want to try is taking a class at a community college. I know that there are several where I live that offer courses in 3D animation. Try to get your parents/advisors at school to help you with this.
And good luck! 3D modeling and animation helped me to express myself a great deal, not unlike programming.
I am a student at the University of Minnesota. I have been working with faculity from the physics and computer science departments on and off over the last 4 years.
Most of the math we have been taught is not really used in computer science. You have some analysis of run times and memory requirements, but so far nothing that required higher mathmatics.
Differental equations have been the most useful to me. They are used in all kinds of different science and analysis of situations (fiscal models are big on them). From what I have seen undergrad CS is an alright intro to the topic, but you do not get to any of the really interesting stuff(iw you need higher level math) until you hit grad school or start working with researchers. Then it is really dependent on who you are working with.
I am currently taking a grad course titled "Intro to Parallel Computing" and the most mathmatics we have used so far is stats to look at an omega network... not really exciting or informative.Good class and learning lots of stuff, but not quite there yet.
I really think that most of the mathmatics that the students should learn are dependent on the topics they are being taught. Discrete math is very important, but was not very challenging to most of the students that took it. Mostly the mathmatics that are needed for a paticular topic should be taught in the class the topics are presented in.
Think along the lines of self contained courses. I know diversity of ones education is important, but sitting through discrete math was a wasted semester. Sets, unions, intersections, and the pigion hole principel are not that hard to understand... and I can not remember when "proving" something was correct was actually useful. (I read a story about how this researcher proved this very short peice of code correct. Three other researchers found MANY errors for how small the code was.) Proving code may have its place but it is not something 99.9% of us are going to be doing with any frequency.
What I would have rather seen is software enginering concepts introduced eailer and the chance to actually USE and explore the ideas presented there./me is still bitter about the poorly run software eng course he took. It was so bad they are totally changing the course.
Anyways:
-diff eqs are good
-software eng good
-discrete math good but we DONT need a whole semester
-get involved with your profs it is one of the best thing you can do!
-The UMN is a great cs school. They listen to their students, have great instructors (we have many that have worked in industry and know their stuff) and is generaly a good place to be (ignoring the code, you are cuddled up next to your sparc station right?)
Ben
I would have loved to just toss the boxes. But I had a job and it was to work with windows boxes. Trust me the small amount of perl I have used I love. I am currently taking a class that is teaching me about the theory of regual expressions. I love that stuff. This was about 3-4 years ago and I did not know about cygwin (was it out, was it free?). I did not smoke back then either... hey look at that I started smoking and learned UNIX, all of a sudden sys admining is easier:)
I agree the perl and unix are powerful and can do just about anything, but my point is that there was an excuse. It was money.
Bad form to reply to self but I totally went and forgot. I designed and implemented a system that used gpg (which looks a great deal like this product, I said looks I am not implying anything other then the functionality is similar) to create a 'secure' file distribution network system via ftp on windows platforms. Durring that project there were a good deal of troubles with not having cli tools.
btw does anyone know how to have a free high quality sshd/sftp server on a windows box?
Ya ya ftp sux, but the data was well enough protected as it sat on the sight (via gpg). The OS was another question.
Thanks man. I guess I should have clairfied(sp?) more in my orginal post. I have done that now. Oh ya btw I have been told many times by people with more experience in the IT industry and paticular applications that something can not be done. I like to show them that it can be it they would just think a bit more creativitly. I love it when things are used for porposes that they were not orginally designed for.
Yup I was talking about the applications that can be called from stuff like cron and at. The reason I brought up windows was when I used to work on those systems I had difficulties finding cli based applications.
The reason command line tools are very useful is for cron jobs. I dont know how many times on a windows machine I wish that there was an command line tool to do something.
If you are using a single drive on either end, getting a faster interconnect will not help you. The drive will be your limiting factor.
If you have a raid array on each side then you would not be limited by the bandwidth off of the drives, but this would be expensive.I have seen some suggestions of getting a firewire HD and just share it between the machines. I think this would be one of your best bets.
Do not worry about the connection speed until you get more/faster spindles.
Software Engineering is about designing and specifying how a software system should work. A good Software Engineering process should be repeatable and reduce the amount of restructuring needing to be done later in the projects life time.
I just put a thin (paper) notebook between my lap and laptop so I don't end up with burnt skin. Keeps my lap nice and warm durring the long cold nights too./me looks around for a girlfriend
The bus and memory speeds are currently the limiting factor in executing code. What he is saying would allow for the extra registers on the CPU to be used as a small fast storage space.The reason this is so good is if you didnt you the registers on board you would have to push the data onto the stack. This means that you have to push data over the bus and into memory, then to get it back you have to pop the data back from memory onto the bus. Two things about that: I that is 2 transphers over the bes that are wasted, and main memory is slower then snot.
So with what Rick has said we aleaviate some bus traffic AND we have faster access to the data that we need. There are ways around this with special software libaries, but then you have to use those compliers/libaries. This would be an addition that all compilers could easly implement and use.
In short no we should concentrate on bus speed or memory speed.
In a light water reactor when a reaction "runs away" the reaction dies, resulting in nothing exploding. The problem with breeder reactors is that when a reaction runs away there is no way to stop in. This is a *very* bad thing.
Breeder reactors are very efficent but you always have to prepare for the worst.
How does this one float peoples boat? School districts buy bare bones computers from Walmart and install Linux on them. The schools get nice cheap machines, with inexpensive software.
The only reason that NAS vendors are using CIFS is that some version of CIFS is bundled into every windows install. If Microsoft included NFS support this problem would not exist. But it is not in Microsoft's best intrests in let go of CIFS. So what is a admin to do? Well there are *lots* of nfs client implementations, but all of the ones I could find are commercial. If there was a free or open source NFS client that ran on windows (heck if it could be integrated into windows) it would be a blow against Microsoft in a big way.
Slashdot Mirror
The current and past trend in databases is to make everything work in a SQL database. As far as commercial relational databases go SQL is the standard. Check out M. Stonebraker, "Inclusion of New Types in Relational Data Base Systems", IEEE Data Engineering Conference, Los Angeles, CA, February 1986. (http://citeseer.nj.nec.com/ncontextsummary/11868/ 52519). This is a paper on how to extend relational databases to handle all sorts of data. It has been used to create Object Relational databases, add in spatial information to databases, and many other types of data. The latests and greatest data type that people are trying to integrate is XML. This has not met with so much sucess when efficently querying the data, but there is a great deal of current research on the subject.
>It seems that they are comparing relational databases. Search engines use proprietary databases which,
Yes and Yes
>among other things, do not allow for live insertion of records, SQL commands, etc.
If they can not insert (bulk load is an insert) how are they going to expand their databases?
>Google probably stores at least 10^9 raw web pages in their cache(that's 10 TB alone)
Google does not store raw web pages. Use google to find the pagerank and pagerate algorithms. Also find the paper on how the google filesystem runs.
After spending literly 30 seconds reading the man page, and a day getting used to it, Ion was the best window manager I have ever used. It was designed to be very efficent, and worked great as a programming interface.
http://modeemi.cs.tut.fi/~tuomov/ion/
The point would be to have a small cluster (5 nodes would do 4 cluster + a login box) so you could learn how to code/admin that type of thing. All of the software is free and the hardware is free to you. So take a spin at learning MPI, OpenMOSIX, or sockets programming!
Currently the top 5 consist of 4 machines that have a Therotical maximum speed (Rmax) the are larger then the 10TFLOPS this machine will have. Then you have to translate that into peak speed which is what matter and what this list uses to rank the machines. Peak will be a good deal less, but this mostly has to do with the way the systems are interconnected and not the machines themselves. Say what you may about the G5 but the interconnect is more important.
There is only one machine in the top 5 that this cluster could beat. The rest of the world has had 6 months to build machines too.
This should be a top 10 machine for sure. Good to see more fast machines being built every day.
Bull Shit. Doing science (be it playing with chemicals or computers) just for the sake of doing science is what has given us MANY of our most useful tools. Penicillin and Lasers were discovered because some scientist was messing around with something and said "Wait, why is that happening?". We also get things that are just part of the journey to destination. This would include things like Velcro.
Besides that what is so wrong about people trying to do things cheaper? More reliable? Because they are interested in them?
If they can make this go as fast as (or close) the theory says, this would be a pretty good platform. Keep in mind that we will be seeing clustering of different types of systems for a long time to come. The more research we do the better we will be at it. In the end it will allow us to be more efficent.
I started out with a copy of truespace a a book. I worked with that for a couple of months. My first job was doing some 3-D modeling and animation for a presentation. I made a little bit of money then "upgraded" to 3DSMAX.
When I say "upgraded" it was an upgrade but nothing that was beyond belief. Yes it was multi-threaded and could do network rendering, there were some other things that it could handle, but it took away productuvity for some time while I learned the new system.
What I would suggest to you is to pick a system that you can afford (or "borrow") and get yourself a book or two. Have some ideas of what you would like to do for a project and learn how to get there.
It takes a lot of time and, for me, a change in how I interacted with the computer to become good enough to claim my work as professional.
One of the other things you might want to try is taking a class at a community college. I know that there are several where I live that offer courses in 3D animation. Try to get your parents/advisors at school to help you with this.
And good luck! 3D modeling and animation helped me to express myself a great deal, not unlike programming.
I am a student at the University of Minnesota. I have been working with faculity from the physics and computer science departments on and off over the last 4 years. Most of the math we have been taught is not really used in computer science. You have some analysis of run times and memory requirements, but so far nothing that required higher mathmatics. Differental equations have been the most useful to me. They are used in all kinds of different science and analysis of situations (fiscal models are big on them). From what I have seen undergrad CS is an alright intro to the topic, but you do not get to any of the really interesting stuff(iw you need higher level math) until you hit grad school or start working with researchers. Then it is really dependent on who you are working with. I am currently taking a grad course titled "Intro to Parallel Computing" and the most mathmatics we have used so far is stats to look at an omega network... not really exciting or informative.Good class and learning lots of stuff, but not quite there yet. I really think that most of the mathmatics that the students should learn are dependent on the topics they are being taught. Discrete math is very important, but was not very challenging to most of the students that took it. Mostly the mathmatics that are needed for a paticular topic should be taught in the class the topics are presented in. Think along the lines of self contained courses. I know diversity of ones education is important, but sitting through discrete math was a wasted semester. Sets, unions, intersections, and the pigion hole principel are not that hard to understand... and I can not remember when "proving" something was correct was actually useful. (I read a story about how this researcher proved this very short peice of code correct. Three other researchers found MANY errors for how small the code was.) Proving code may have its place but it is not something 99.9% of us are going to be doing with any frequency. What I would have rather seen is software enginering concepts introduced eailer and the chance to actually USE and explore the ideas presented there. /me is still bitter about the poorly run software eng course he took. It was so bad they are totally changing the course.
Anyways:
-diff eqs are good
-software eng good
-discrete math good but we DONT need a whole semester
-get involved with your profs it is one of the best thing you can do!
-The UMN is a great cs school. They listen to their students, have great instructors (we have many that have worked in industry and know their stuff) and is generaly a good place to be (ignoring the code, you are cuddled up next to your sparc station right?)
Ben
I would have loved to just toss the boxes. But I had a job and it was to work with windows boxes. Trust me the small amount of perl I have used I love. I am currently taking a class that is teaching me about the theory of regual expressions. I love that stuff. This was about 3-4 years ago and I did not know about cygwin (was it out, was it free?). I did not smoke back then either... hey look at that I started smoking and learned UNIX, all of a sudden sys admining is easier :)
I agree the perl and unix are powerful and can do just about anything, but my point is that there was an excuse. It was money.
Bad form to reply to self but I totally went and forgot. I designed and implemented a system that used gpg (which looks a great deal like this product, I said looks I am not implying anything other then the functionality is similar) to create a 'secure' file distribution network system via ftp on windows platforms. Durring that project there were a good deal of troubles with not having cli tools. btw does anyone know how to have a free high quality sshd/sftp server on a windows box? Ya ya ftp sux, but the data was well enough protected as it sat on the sight (via gpg). The OS was another question.
Thanks man. I guess I should have clairfied(sp?) more in my orginal post. I have done that now. Oh ya btw I have been told many times by people with more experience in the IT industry and paticular applications that something can not be done. I like to show them that it can be it they would just think a bit more creativitly. I love it when things are used for porposes that they were not orginally designed for.
Yup I was talking about the applications that can be called from stuff like cron and at. The reason I brought up windows was when I used to work on those systems I had difficulties finding cli based applications.
The reason command line tools are very useful is for cron jobs. I dont know how many times on a windows machine I wish that there was an command line tool to do something.
If you are using a single drive on either end, getting a faster interconnect will not help you. The drive will be your limiting factor.
If you have a raid array on each side then you would not be limited by the bandwidth off of the drives, but this would be expensive.I have seen some suggestions of getting a firewire HD and just share it between the machines. I think this would be one of your best bets.
Do not worry about the connection speed until you get more/faster spindles.
Software Engineering is about designing and specifying how a software system should work. A good Software Engineering process should be repeatable and reduce the amount of restructuring needing to be done later in the projects life time.
I just put a thin (paper) notebook between my lap and laptop so I don't end up with burnt skin. Keeps my lap nice and warm durring the long cold nights too. /me looks around for a girlfriend
The bus and memory speeds are currently the limiting factor in executing code. What he is saying would allow for the extra registers on the CPU to be used as a small fast storage space.The reason this is so good is if you didnt you the registers on board you would have to push the data onto the stack. This means that you have to push data over the bus and into memory, then to get it back you have to pop the data back from memory onto the bus. Two things about that: I that is 2 transphers over the bes that are wasted, and main memory is slower then snot.
So with what Rick has said we aleaviate some bus traffic AND we have faster access to the data that we need. There are ways around this with special software libaries, but then you have to use those compliers/libaries. This would be an addition that all compilers could easly implement and use.
In short no we should concentrate on bus speed or memory speed.
In a light water reactor when a reaction "runs away" the reaction dies, resulting in nothing exploding. The problem with breeder reactors is that when a reaction runs away there is no way to stop in. This is a *very* bad thing. Breeder reactors are very efficent but you always have to prepare for the worst.
How does this one float peoples boat? School districts buy bare bones computers from Walmart and install Linux on them. The schools get nice cheap machines, with inexpensive software.
The only reason that NAS vendors are using CIFS is that some version of CIFS is bundled into every windows install. If Microsoft included NFS support this problem would not exist. But it is not in Microsoft's best intrests in let go of CIFS. So what is a admin to do? Well there are *lots* of nfs client implementations, but all of the ones I could find are commercial. If there was a free or open source NFS client that ran on windows (heck if it could be integrated into windows) it would be a blow against Microsoft in a big way.