Abstract: Example implementations relate to maintaining service availability. According to an example, a distributed system including multiple nodes establishes a number of the nodes to operate as voters in a quorum evaluation process for a service supported by the nodes. A first proper subset of the nodes is identified to operate as the voters and the remainder of the nodes do not vote in the quorum evaluation process. Responsive to detecting, by a voter, existence of a failure that results in the voter being part of a second proper subset of the nodes and that calls for a quorum evaluation, the voter determines whether the second proper subset of nodes has a quorum by initiating the quorum evaluation process. When the determining is affirmative, the second proper subset continues to support the service; otherwise, the second proper subset discontinues to make progress on behalf of the service.

Abstract: Techniques and mechanisms for establishing quorums on an object-by-object basis; e.g., by associating quorum schemes, quorum-eligibility groups and/or quorum rules with objects on an object-by-object basis. For example, a certain quorum scheme may be associated with one object in a system and a different quorum scheme may be associated with another object in the system. A management system may be configured to be part of a quorum-eligibility group of an object based on any of a variety of factors, including any of: the number of management systems in the system, the quorum rule to be employed; whether an instance of the object is stored in a same fault domain of at least one of the management systems that are members of the quorum scheme; and the desire to maintain high availability of the object in response to various failure scenarios.

Abstract: A system for managing distributed storage includes a collection of one or more compute sites that are connected and that perform a joint computation. A witness node is connected to the one or more compute sites. A membership service allows a change in collection membership, in response to a failure, to continue the joint computation without causing data inconsistency or corruption. If the membership service is unable to reach a decision on the change in collection membership in response to the failure, the witness node controls the decision on the change in collection membership to continue the joint computation. A distributed storage system provided by the system described herein may advantageously meet consistency guarantees and maximize data access even when individual compute sites and/or their interconnects fail.