In the B-Virtual Distributed Storage System, the way data is distributed across the available storage nodes is policy controlled. For each data set, you can configure the minimum number of nodes that need to be included in the data spread and the number of simultaneous disk failures in this spread that the system needs to be able to survive. Based on these parameters, the distributed storage system select the disks in storage nodes such that data is spread as wide as possible across the available nodes and their locations. This guarantees that the impact of a component failure is minimal and never jeopardizes data availability.
For instance you could configure the storage system to spread a dataset over 16 storage nodes in the DSS. The availability requirement is to survive up to 4 simultaneous node failures. In that case, the DSS will spread incoming data redundantly over a selection of 16 disks in different storage nodes that are wide apart. Any 4 storage nodes out of the 16 can fail without impacting the data availability. The DSS can retrieve your data from any arbitrary subset of 12 nodes out of the 16.
In the event of a storage node or disk failure, the B-Virtual Distributed Storage System will automatically reconfigure itself to store new incoming data according to an alternative spread of available storage nodes, according to the policy. Thus if for instance node 5 in the spread of 16 in our previous example would fail, then a new spread of 16 available nodes that does not include node 5 will be selected. Newly incoming data is immediately written to the new spread of 16 nodes, thus the availability policy is met immediately for new incoming data.
Meanwhile, in the background, storage maintenance agents regenerate the data of the failed disk on available disk capacity in another storage node. You can configure storage maintenance agents to run permanently or only during maintenance windows.
At any time, storage capacity can be added to the B-Virtual DSS. When the system reorganizes the data spread because of a node failure or when a storage node has filled up, this added storage capacity will automatically be included in newly generated spreads. The added capacity is thus immediately available to the storage system without any reconfiguration.