Help in setting up a cluster
Yazeed Alsaif
Hello,
This is my first time posting here, I think I am in the right place, if not I am sorry.
um, Here is my situation.
I am in middle of my final year project (Undergrad), anyway my project is on Linux Clustering and I am asked to demonstrate the effectiveness of a cluster practically, I have done my thesis but I am stuck on the practical side.
what I want and looking for is:
1- I want to test the cluster by running a software that can take advantage of other machines resources and converting songs from one extension to another in a short time, and then I will compare that to if the conversion process was run on one machine.
2- I want to run a web server
on the cluster, so if one machine fails, the website will still be there. (for fault tolerance demonstration)
Note: I do NOT anyone to do it for me, I want to do it myself. All I want is guidelines and perhaps some best practices.
thanks guys.
Yazeed.
Anatoly Molchanov
Gerald Wrong
The webserver failover is a slightly different topic from distributed/cluster computing IMO. One easy thing to do would be to make an image whereby every machine had the same webserver and content, and use some kind of load balancing/redirection.
Good Luck!
Pedro
but I don't test it yet
http://www.rocksclusters.org/
On Wed, Feb 20, 2013 at 10:37 PM, Gerald Wrong <psillym...@gmail.com> wrote:
>
> Look into Mosix for simple heterogenous clustering of linux boxes, but you have to comfortable rolling your own kernels unless somethings changed. Last time I was a student, you could get Mosix for free as a student.
>
> The webserver failover is a slightly different topic from distributed/cluster computing IMO. One easy thing to do would be to make an image whereby every machine had the same webserver and content, and use some kind of load balancing/redirection.
>
> Good Luck!
>
>
> On Wed, Feb 20, 2013 at 3:58 PM, Anatoly Molchanov <ykd...@gmail.com> wrote:
>>
>> Do you want use beowulf? Or you interested in clustered web-server issue?
>>
>>
>> 2013/2/21 Yazeed Alsaif <saify...@gmail.com>
>>>
>>> Hello,
>>>
>>> This is my first time posting here, I think I am in the right place, if not I am sorry.
>>>
>>> um, Here is my situation.
>>> I am in middle of my final year project (Undergrad), anyway my project is on Linux Clustering and I am asked to demonstrate the effectiveness of a cluster practically, I have done my thesis but I am stuck on the practical side.
>>> what I want and looking for is:
>>>
>>> 1- I want to test the cluster by running a software that can take advantage of other machines resources and converting songs from one extension to another in a short time, and then I will compare that to if the conversion process was run on one machine.
>>>
>>> Note: I do NOT anyone to do it for me, I want to do it myself. All I want is guidelines and perhaps some best practices.
>>> thanks guys.
>>>
>>> Yazeed.
>>
>>
>
To UNSUBSCRIBE, email to debian-beow...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listm...@lists.debian.org
Archive: http://lists.debian.org/CABr7qmSXukLBEoehMFtaJkmh...@mail.gmail.com
Geoffrey Jacobs
Look into Mosix for simple heterogenous clustering of linux boxes, but you have to comfortable rolling your own kernels unless somethings changed. 锟絃ast time I was a student, you could get Mosix for free as a student.
The webserver failover is a slightly different topic from distributed/cluster computing IMO. 锟絆ne easy thing to do would be to make an image whereby every machine had the same webserver and content, and use some kind of load balancing/redirection.
Good Luck!
On Wed, Feb 20, 2013 at 3:58 PM, Anatoly Molchanov <ykd...@gmail.com> wrote:
Do you want use beowulf? Or you interested in clustered web-server issue?
Hello,
This is my first time posting here, I think I am in the right place, if not I am sorry.
um, Here is my situation.
I am in middle of my final year project (Undergrad), anyway my project is on Linux Clustering and I am asked to demonstrate the effectiveness of a cluster practically, I have done my thesis but I am stuck on the practical side.
what I want and looking for is:
1- I want to test the cluster by running a software that can take advantage of other machines resources and converting songs from one extension to another in a short time, and then I will compare that to if the conversion process was run on one machine.
Mosix is pretty much dead now, IIRC. You could try Kerrighed or OpenSSI.2- I want to run a web server
锟給n the cluster, so if one machine fails, the website will still be there. (for fault tolerance demonstration)
Note: I do NOT锟絘nyone to do it for me, I want to do it myself. All I want is guidelines and perhaps some best practices.
thanks guys.
Yazeed.
Yetkin Öztürk
I recommend you to try pelicanhpc http://pareto.uab.es/mcreel/PelicanHPC/ it does not need installation, it can boot from live disk and make remote nodes boot from network. Its a very practical way to see how a cluster look like and to try some mpi coding.(also based on debian)
Yetkin Ozturk
21 Şubat 2013 Perşembe tarihinde Geoffrey Jacobs <gdja...@gmail.com> adlı kullanıcı şöyle yazdı:
> On 02/20/2013 03:37 PM, Gerald Wrong wrote:
>
>
> The webserver failover is a slightly different topic from distributed/cluster computing IMO. 燨ne easy thing to do would be to make an image whereby every machine had the same webserver and content, and use some kind of load balancing/redirection.
>
> Good Luck!
>
> On Wed, Feb 20, 2013 at 3:58 PM, Anatoly Molchanov <ykd...@gmail.com> wrote:
>>
>> Do you want use beowulf? Or you interested in clustered web-server issue?
>>
>>
>> 2013/2/21 Yazeed Alsaif <saify...@gmail.com>
>>>
>>> Hello,
>>>
>>> This is my first time posting here, I think I am in the right place, if not I am sorry.
>>>
>>> um, Here is my situation.
>>> I am in middle of my final year project (Undergrad), anyway my project is on Linux Clustering and I am asked to demonstrate the effectiveness of a cluster practically, I have done my thesis but I am stuck on the practical side.
>>> what I want and looking for is:
>>>
>>> 1- I want to test the cluster by running a software that can take advantage of other machines resources and converting songs from one extension to another in a short time, and then I will compare that to if the conversion process was run on one machine.
>>>
>>>
>>> Note: I do NOT燼nyone to do it for me, I want to do it myself. All I want is guidelines and perhaps some best practices.
Stéphan Gorget
your two different projects :
1 - (converting song) - use a scheduler such as torque / slurm-llnl
(debian package name) / sun grid engine (the name might have change)
and use those to distribute the different jobs across the different
nodes. You might need to have a way to authenticate user across all
the node + a filesystem such as NFS to share the files.
2 - (load balancer/high availability) - you might need to consider
different solutions that will be able to fail over such as
heartbeat/pacemaker or corosync/pacemaker, if you want to do load
balancer you might need to look at haproxy or some other high
availability solutions.
Stéphan
To UNSUBSCRIBE, email to debian-beow...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listm...@lists.debian.org