SSIC gives you a Single System Image Cluster. SSIC distributes jobs across the cluster so that no one system will be overly burdened running a particular process.http://sourceforge.net/projects/ssic-linux/
Dual-SSIC gives you 2 SSI clusters running side by side. Call it a cluster mirror if you will. Dual-SSIC will be mirrored cluster kinda like hard drives can be mirrored with raid 1.
Details:
1 node configuration:
One node could be used, but that's kinda useless in a cluster environment -- we'll leave this one for testing only or for when one of your 2 nodes dies (in a 2 node configuration).
2 node configuration:
Master node gets set up. The slave node boots and synchronizes its filesystem with the master. Once the filesystem is up to date then all running processes on the Master node is duplicated to the slave node. The processes on the slave node will continue to run as normal and will keep in step with the master node.
The stdin may be tied to the master stdin as per the above URL. Actions in processes will update the slave filesystem. A filesystem synch process may monitor the filesystems to make sure it is in synch with the master? If the slave node encounters an error in the filesystem or process listing the error will be corrected and a bug report will be generated for the Adminstrator to send in.
In summary we have a mirrored filesystem between the master and slave nodes, and we have all processes mirrored as well. All processes on the slave node run lock step with the master and affect its filesystem the same way it affects the master. Only stdout from the master node is sent back to the user. And only one official IP address for the whole cluster using LVS (http://www.linuxvirtualserver.org/HighAvailability.html).
More than 2 node configuration:
In brief, adding more than 2 nodes will increase the overall speed of the cluster by spreading jobs out to nodes that are not as busy.
Same as the 2 node configuration, except that we have cluster A (master) and cluster B (slave). All new nodes added will be added to either cluster A or cluster B depending on which cluster needs more CPU power (the decision will be made automatically by the overall cluster just as SSIC decides a process needs to migrate). There will still be only 2 copies of any process at any given time, there will be only one on cluster A and one on cluster B.
As new nodes are added it is decided to which cluster (A or B) they need to be added to, their filesystem is brought up to date, and processes are migrated to it as SSIC determines necessary.
Random notes:
The filesystem will be identical on all cluster nodes.
One node is the master and all other nodes are slaves.