Re: [SSI-devel] nodes

[Keith W.B. <ke...@co...>]

I had looked at using   HA + LVM + LTSP or LVM+DR  but the shared  
memory and the way processes are handled are a huge feature.  I also  
read some comparisons of a running install of a few SSI  clusters  
types; openssi again handled exactly the way we want to use it.

I've spoken with someone else and he recommended the HA LVM LTSP,  
what are your thoughts?

apache diskless cluster very few files being used, all using a  
database on some other servers not in the cluster  100 to 200 servers  
to start,  500 is my goal

Thanks for all your help.

-Keith


On Dec 7, 2006, at 2:19 PM, Keith W. Bacon wrote:

> Thank you for your replies, honestly I'm looking to setup 500 or  
> more nodes. 100 to 200 initally
>
> I'm having a difficult time getting this to recompile correctly for  
> Debian sarge and Debian testing.  I pulled down the cvs with same  
> results.  Any assistance would be much appreciated!
>
> Thank you,
>
> -Keith
>
>
> On Dec 7, 2006, at 2:15 PM, Laura Ramirez wrote:
>
>> I believe the openmosix code wont compile if its over 125.
>>
>> laura
>>
>> Brian J. Watson wrote:
>>> Keith W.Bacon wrote:
>>>> Hi everyone!
>>>>
>>>> The number of nodes on the site says 125
>>>>
>>>> Is this up to date?  I read about using procs and the number of  
>>>> nodes  is thousands.
>>>>
>>>> Openssi has been the perfect clustering software and I like to  
>>>> know  if I'll be able to scale to the moon or in the next version?
>>>>
>>>> I will be setting up a diskless openssi cluster of 100  apache  
>>>> servers.
>>> Hi Keith,
>>> IIRC, there is a #define in include/linux/cluster.h (or one of  
>>> the other SSI header files) that sets the limit on the number of  
>>> nodes. By changing this and recompiling the kernel, you'll be  
>>> able to add more nodes. It is possible that increasing this  
>>> number could trigger some bugs, but I don't know that it would.  
>>> It might just be a testing effort to see if it works correctly  
>>> with say 150 or 200 nodes.
>>> I know that there are some scaling problems that start kicking in  
>>> at some point, which would need to be fixed by re-architecting  
>>> some algorithms, but I think it would require several hundred  
>>> nodes to trigger these scaling problems.
>>> If you're interested in testing OpenSSI with more nodes, I'm sure  
>>> the developers on this list would be interested in your results.
>>> Regards,
>>> Brian
>>> -------------------------------------------------------------------- 
>>> -----
>>> Take Surveys. Earn Cash. Influence the Future of IT
>>> Join SourceForge.net's Techsay panel and you'll get the chance to  
>>> share your
>>> opinions on IT & business topics through brief surveys - and earn  
>>> cash
>>> http://www.techsay.com/default.php? 
>>> page=join.php&p=sourceforge&CID=DEVDEV
>>> _______________________________________________
>>> ssic-linux-devel mailing list
>>> ssi...@li...
>>> https://lists.sourceforge.net/lists/listinfo/ssic-linux-devel
>