Yes, agreed. In my case, I already have fibre channel storage set up now. Each node can see the storage as it's own so I can avoid things such as NFS for example. I like the idea of network boots for two reasons. One is that I have a drive in each blade which costs power and yet, I still have to add drives into the central storage. Seems that I could save money and power by not having any drives in each blade/server. I'd prefer that. Second is that right now, when I need a new node, I have to install a new server, configure the server rather heavily etc, and in the end, that server and my time is not being used very effectively.. it seems. Most of the things I want to run seem to allow for central storage. I am running email, web and sql services for the most part, other things might be on their own or their own clusters. Web servers can share data, mail servers and from what I've seen, sql servers can too. So, when I need more resources, it would be a simple matter of configuring the net boot options and firing up another blade and I'm done. Of course there are other aspects but I'm talking about the basics. Am I missing something? Mike On Thu, 26 Oct 2006 15:40:11 -0400, Patton, Matthew F, CTR, OSD-PA&E wrote: > Classification: UNCLASSIFIED > > before getting too mired in minutia, read/reread all the howto etc on > netbooting. things like /usr lend themselves nicely to read-only NFS (or > GFS) mounts. / (root) should be writable by just the host it applies to. So > GFS is again, not particularly useful IMO. mailspools and common > html/jsp/applet sort of directories are candidates for GFS. Before you go > with clustering, might want to think hard whether it makes sense to go to > all that bother. a load balancer (eg. piranah or commercial) and machines > mounting NFS may accomplish what you are really after. -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
