Block storage in OpenStack code named Cinder has resolved a few major issues that exist in the former OpenStack releases. Instead of counting on ephemeral storage, creating a persistent storage volume will prevent the deletion of data when the instance is terminated. Cinder API offers a full set of block storage management interface that allows creating, removing, resizing, attaching, and detaching volumes to instances. This is also the case for Sahara. It is possible to create and attach volumes to instances associated to a node group during cluster provisioning. This is a very fruitful feature from a reliability perspective since a Hadoop cluster relies on its HDFS.

Cinder includes a scheduler that schedules the location of new volumes based on a filter policy. Based on the highest weight backend, the cinder-scheduler will assign to it the creation of a new volume. From a performance perspective, the block storage service in OpenStack allows to define explicitly...