CCL:Largest Linux Cluster? -- ide

alvin at Maggie.Linux-Consulting.com alvin at Maggie.Linux-Consulting.com
Thu Jan 24 11:06:46 PST 2002


hi ya craig

On Thu, 24 Jan 2002, Craig Tierney wrote:

> Saying that 'they use no local disk, opting for a huge disk farm' is
> not exactly correct.  All nodes are diskless, and they do have a huge disk
> farm.  The san is for data storage, user codes, etc.  The nodes are diskless
> because they need to be able to switch parts the machine from classified to
> non-classified depending on the situation.  They cannot have any storage locally
> to accomplish this.  It doesn't take much disk to boot 1000 nodes remotely. 
> All common files are shared, and only some system specific things such as /etc
> are generated unique to each node.  I think (not positive) that /var and /tmp
> may be small ram disks, but I don't remember exactly.
> 
> Does hdparm work with SCSI?  Does it really accomplish much?  I didn't think
> it did.  They SAN is scsi, so I don't think they are running hdparm to optimize it.

hdparm is for ide.. ( am thinking on ide-based raid5...

for the other responder...

and yes...fibrechannel is good too, but again, any dual hosted controllers
for it too ??? 

have fun linuxing
alvin
http://www.Linux-1U.net .. 8x 200GB IDE disks -->> 1.6TeraByte 1U Raid5 ..




More information about the Beowulf mailing list