A Cluster over Ethernet with Open-E JovianDSS is built of two nodes. Each node has its own hard drives attached directly to the node. The solution allows for using a commodity server that is made up of only SATA drives.
All hard drives on one node are shared with the other node, and the whole configuration works almost like a regular cross-linked cluster connection. However, instead of a direct connection of HDDs to both nodes, they are directly connected to one node and (after internal conversion of the protocol) indirectly served over Ethernet to the other node. Each node can see all the hard drives in the cluster. The other node’s drives are marked as ‘Remote Disks’.
In order to achieve cluster redundancy users are required to create mirrors between local and Remote Disks.
Since the connection of cluster communication and data mirroring between nodes works over Ethernet, the nodes might be located a very long distance apart (as a metro storage cluster or a stretched metro storage cluster). It can be 50 miles (80 km) in case of point to point fibre optic connection, or even more when using an additional switch between nodes - provided that network latency will not exceed 5 ms.
This active-active dual node cluster provides load balance for SMB, NFS, FC, and iSCSI protocols due to active pools on both nodes. This way, in case one node fails, the other one takes over its services and the pools from the first node are immediately activated on the other node. By balancing the load on both nodes, they can be utilized to their fullest potential.
Other possible configurations: