What can be done with old computer equipment?

[philocrazy]

franznietzsche

i would. twelve times 200 mhz gives me an extra 2.4 ghz. Very useful because it would give me a dedicated system for my work. Unfortunately, straving student, i don't have $600, and i doubt I'm going to be able to convince the uni to cover it.

-------------------------------------
buy them! and youll still get 200mhz not to mention the old buses that run at hot
speeds ,franznietzsche you can find more machines of such speeds to add up to greater speeds say 300ghz
guys the linux guru franznietzsche

 

[dduardo]

You can't just do 12 times 200Mhz. There are things like overhead which will reduce the maximum speed achievable. You'll be lucky to get 1.2ghz from all twelve computers.

 

[franznietzsche]

buy them! and youll still get 200mhz not to mention the old buses that run at hot
speeds ,franznietzsche you can find more machines of such speeds to add up to greater speeds say 300ghz
guys the linux guru franznietzsche

I sense ridicule, coming from someone who doesn't know how to use quote tags.

You can't just do 12 times 200Mhz. There are things like overhead which will reduce the maximum speed achievable. You'll be lucky to get 1.2ghz from all twelve computers.

I will admit my own ignoarance in hardware setups. What specifically would be limiting the number of cycles you could get?

 

[dduardo]

Some of the factors that cause overhead:

1. The speed of the network.

In a non-clustered configuration a single computer a stick of PC100 can be read by the cpu at 800MB/s. With PC133 it is 1GB/s. Therefore, if you using a 100baseT network your bottlenecking the speed at which one cpu can get data from other node on the cluster. This is why high end clusters use fiber optic.

2. The software managing the sharing of the load.

Managing data over multiple nodes is very complicated. You need to keep track of all the processes going on at all times. Syncing processes/memory is one of the major design issues encoutered when writing clustering management software. If not designed properly, you'll get terrible performance.

3. The program you want to run.

Not all programs can be split. Depending on how the software was programmed will dictate how well the process can be split.

 

[Tom McCurdy]

I am getting lost in the discussion are people debating on whether or not a 12 Power Computing mac beowulf cluster would be worth setting up?

 

[franznietzsche]

I am getting lost in the discussion are people debating on whether or not a 12 Power Computing mac beowulf cluster would be worth setting up?

Philocrazy might, but I'm just asking about clusters in general.

 

[dduardo]

Basically, if all you're after is learning how clustering works then go for it. If you're after performance, don't even bother.

 

[Wardw]

Who is paying for all the electrons needed to power this daughter of Frankenstein??
:-)

 

[Tom McCurdy]

Hehehe Mom and Dad

 

[franznietzsche]

Hehehe Mom and Dad

always good.

I just got another hard drive so that i can add two more operating systems to my computer, which they were kind enough to cover for me.

 

[FulhamFan3]

I think it's more work than it's worth to make a cluster with that unless it's simply to learn how clusters work. However 2 other uses comes to mind.

1.) If the hard drives are still useful you can set them up as file servers.

2.) Since one of the most popular games in the world was made for a machine of those specs you can make a big Starcraft LAN and invite your buddies.

 

[franznietzsche]

2.) Since one of the most popular games in the world was made for a machine of those specs you can make a big Starcraft LAN and invite your buddies.

*shakes head*

People still play starcraft?

Its not that good a game, its really not.

 

[FulhamFan3]

*shakes head*

People still play starcraft?

Its not that good a game, its really not.

Go to south korea and try to find someone that doesn't play starcraft. Is it really starcraft you don't like or the strategy game genre in general?

 

[franznietzsche]

Go to south korea and try to find someone that doesn't play starcraft. Is it really starcraft you don't like or the strategy game genre in general?

I don't consider starcraft strategy, its all simply micromanagement.

I love strategy games, real strategy games anyway. I'm a serious Total war addict for one. Any game in the starcraft engine (including the age of ------ series) is not really strategy though. Or at most, its unrealistic strategy. I like the total war games that really allow for some strategic innovation in battlefield tactics (well, not everyone innovates, but its possible at least).

 

[Tom McCurdy]

I think it's more work than it's worth to make a cluster with that unless it's simply to learn how clusters work. However 2 other uses comes to mind.

1.) If the hard drives are still useful you can set them up as file servers.

2.) Since one of the most popular games in the world was made for a machine of those specs you can make a big Starcraft LAN and invite your buddies.

Would it be worth all that effort for a whole bunch of 2 gig hard drives

 

[Tom McCurdy]

Also if I were to sell the computers would I be able to make more money if I took off all the old mac OS and loaded say gentoo onto them all?

 

[dduardo]

hahahaha. You expect to compile gentoo on those old machines. You'll be luckly if you can finish by the end of next year.

personally, I don't think you'll get much for the complete system. Your better off selling the components individually.

 

[Christophe]

hahahaha. You expect to compile gentoo on those old machines. You'll be luckly if you can finish by the end of next year.

personally, I don't think you'll get much for the complete system. Your better off selling the components individually.

How can you compile gentoo ?

-Christophe Messiana

 

[dduardo]

Gentoo is a source based distribution. You compile the whole system optimized to your processor and specific needs. To learn more read the gentoo handbook:

http://www.gentoo.org/doc/en/handbook/handbook-x86.xml

 

[Pfft]

If you had a large number of microprocessors in series and a large number of identical problems, you could build a system to step the problems through with a gain in processing time. For example if a single data set can be run through a single processor in ten seconds, break the problem into ten sections each requiring one second in a dedicated processor. To run one set through the series of ten processors still requires ten seconds, but if you keep feeding data in, pretty soon you will be getting answers output every one second.

The FAA developed such a system with the assistance of MIT back in the 1980s using about twenty series connected microprocessors to predict air traffic. The goal was to schedule takeoff times so that the aircraft did not have to orbit around their destination waiting for a opening to land. It was called the Enhanced Traffic Management System (ETMS).

Parallel connected microprocessors are more limited in utility unless you are dealing with irrational numbers a thousand digits long. Modern ciphers use non-repeating series of numbers such as pi to encrypt messages. Since the cipher cracker does not know which irrational number has been used and what portion of the numerical sequence to use, a square matrix of processors would be useful. The X-axis could step the numerical series through while the Y-axis could step the message through and look for repetitive sequences.