Clustering On *BSD? 1
Lonesome asks: "I'm looking at clustering several Free/Open BSD or Solaris servers in a production environment for high availability.
i.e. the machines are linked together, for all purposes look like one machine and the clustering software takes care of keeping the data on the machines in sync (they'll be handling file [including NFS] and mail services) so that if one of them falls over the others will still provide the services, users won't see much difference and minimal data will be lost. All I have been able to find in the way of clustering tools is either in early development or vastly expensive. Does anyone know of a good, preferably BSD/GPL license, solution?"
Does anyone know of a good, preferably BSD/GPL li (Score:1)
On the front end, you'll want a multi-layer ethernet switch that will do server load balancing based on a round-robin or weighted round-robin strategy (http://www.extremenetworks.com or http://www.foundrynet.com), decentralized process migration that is both automatic and transparent (http://www.mosix.org) for your cluster of servers and you'll probably want to set up a robust file server cluster (think scsi, raid and http://oss.missioncriticallinux.com/kimberlite) to serve as a common filestore for you're apps...
I'm not sure what you consider expensive, but if you are putting together a cluster with 16 to 32 decently powered nodes (plus a filestore like kimberlite) the ethernet switch only looks expensive by itself. Present the price of the ethernet switch (which is probably the part of the equation that will make you choke) on a per node basis to management. You'll probably be able to get a larger cluster going 'cause the phb's will see an 'economy of scale' and an upgrade path where each new node looks increasingly less expensive.