Abstract: An asynchronous state machine replication solution in a system of replicas includes executing multiple instances of a consensus protocol, referred to as leader-based views (LBVs) in each replica, where each replica is a leader participant in one of the LBV instances. Each replica drives a decision based on the consensus being reached among the LBV instances, rather than relying the expiration of timers and view changes to drive progress.

Abstract: A method of managing a transaction in a control plane executing on a computing system that manages a plurality of services includes: receiving, at the control plane from a client, a plurality of first requests for at least one target service of the plurality of services, each of the plurality of first requests including a transaction indicator identifying the transaction; executing at least one handler of the at least one target service to process the plurality of first requests; receiving, at the control plane from the client, a commit request for the transaction coordinator service, the commit request including an instruction to commit the transaction; and executing a handler of the transaction coordinator service to process the commit request and notify the at least one target service of a status of the commit request.

Abstract: A replicated service comprises N replicas deployed on compute nodes of a computer network, wherein the N replicas are each configured to vote on a proposed transaction output by a leader of the N replicas and certify the proposed transaction upon receiving qr*N first votes, where qr is a fractional value between 0 and 1 that represents a quorum required for certification. A method of approving a transaction in the replicated service includes receiving certifications from the N replicas, determining whether or not the certifications are received from at least qr*N replicas during a time period equal to 2*?, where ? represents a network delay between two compute nodes of the computer network, and transmitting an approval of the transaction to the replicas for recording by the replicas upon determining that at least qr*N certifications have been received at the end of the time period equal to 2*?.

Abstract: A replicated service comprises N replicas deployed on compute nodes of a computer network, wherein the replicas are each configured to vote on a proposed transaction output by a leader of the replicas and certify the proposed transaction upon receiving qr*N first votes, where qr is a fractional value between 0 and 1 that represents a quorum required for certification. A method of approving a transaction in the replicated service includes receiving certifications from the replicas, and transmitting an approval of the transaction to the replicas for recording: (i) upon determining that at least qc*N certifications have been received, where qc is a fractional value between 0 and 1 that represents a quorum required for transaction approval and qc>qr, or (ii) upon determining that at least qr*N certifications have been received at the end of the time period equal to 2*?, where ? represents a network delay.

Abstract: A replicated service comprises N replicas deployed on compute nodes of a computer network, wherein upon receiving qr*N first votes from other replicas on a proposed transaction by a leader of the N replicas, each of the N replicas certifies the proposed transaction to a client of the replicated service, where qr is a fractional value between 0 and 1 that represents a quorum required for certification. A method of approving a transaction in the replicated service includes receiving the certifications from the N replicas, determining whether or not the certifications are received from at least qc*N replicas, where qc is a fractional value between 0 and 1 that represents a quorum required for transaction approval and qc>qr, and transmitting an approval of the transaction to the replicas for recording by the replicas upon determining that the certifications have been received from at least qc*N replicas.

Abstract: A log unit provides a shared log for recording updates on data objects. Garbage collection is performed locally and in-place by the log unit. In a marking portion of the garbage collection process, the log unit identifies and marks log entries that record mergeable updates. In a deallocation portion of the process, the log unit merges one or more mergeable log entries and deallocates at least portions of the merged log entries.

Abstract: Techniques for implementing dynamic timeout-based fault detection in a distributed system are provided. In one set of embodiments, a node of the distributed system can set a timeout interval to a minimum value and transmit poll messages to other nodes in the distributed system. The node can further wait for acknowledgement messages from all of the other nodes, where the acknowledgement messages are responsive to the poll messages, and can check whether it has received the acknowledgement messages from all of the other nodes within the timeout interval. If the node has failed to receive an acknowledgement message from at least one of the other nodes within the timeout interval and if the timeout interval is less than a maximum value, the node can increment the timeout interval by a delta value and can repeat the setting, the transmitting, the waiting, and the checking steps.

Abstract: Techniques for implementing linear view-change in a Byzantine Fault Tolerant (BFT) protocol running on a distributed system comprising n replicas are provided. According to one set of embodiments, at a time of performing a view-change from a current view number v to a new view number v+1, a replica in the n replicas corresponding to a new proposer for new view number v+1 can generate a PREPARE message comprising a single COMMIT certificate, where the single COMMIT certificate is the highest COMMIT certificate the new proposer is aware of. The new proposer can then transmit the PREPARE message with the single COMMIT certificate to all other replicas in the n replicas.

Abstract: Techniques for implementing linear view-change with optimistic responsiveness in a BFT protocol running on a distributed system comprising n replicas are provided. According to one set of embodiments, the replicas can execute, during a view v of the BFT protocol, a first voting round comprising communicating instances of a first type of COMMIT certificate among the replicas. Further, when 2f+1 instances of the first type of COMMIT certificate associated with view v have been received by the replicas, the replicas can execute a second voting round comprising communicating instances of a second type of COMMIT certificate among the replicas. If 2f+1 instances of the second type of COMMIT certificate associated with view v are not received by the replicas within a predetermined timeout period, a view change can be initiated from view v to a view v+1.

Abstract: A log unit provides a shared log for recording updates on data objects. Garbage collection is performed locally and in-place by the log unit. In a marking portion of the garbage collection process, the log unit identifies and marks log entries that record mergeable updates. In a deallocation portion of the process, the log unit merges one or more mergeable log entries and deallocates at least portions of the merged log entries.

Abstract: Techniques for implementing linear view-change in a Byzantine Fault Tolerant (BFT) protocol running on a distributed system comprising n replicas are provided. According to one set of embodiments, at a time of performing a view-change from a current view number v to a new view number v+1, a replica in the n replicas corresponding to a new proposer for new view number v+1 can generate a PREPARE message comprising a single COMMIT certificate, where the single COMMIT certificate is the highest COMMIT certificate the new proposer is aware of. The new proposer can then transmit the PREPARE message with the single COMMIT certificate to all other replicas in the n replicas.

Abstract: A log unit provides a shared log for recording updates on data objects. Garbage collection is performed locally and in-place by the log unit. In a marking portion of the garbage collection process, the log unit identifies and marks log entries that record supersedable updates. In a deallocation portion of the process, the log unit deallocates at least portions of the marked log entries that contain supersedable updates.

Abstract: Techniques for implementing dynamic timeout-based fault detection in a distributed system are provided. In one set of embodiments, a node of the distributed system can set a timeout interval to a minimum value and transmit poll messages to other nodes in the distributed system. The node can further wait for acknowledgement messages from all of the other nodes, where the acknowledgement messages are responsive to the poll messages, and can check whether it has received the acknowledgement messages from all of the other nodes within the timeout interval. If the node has failed to receive an acknowledgement message from at least one of the other nodes within the timeout interval and if the timeout interval is less than a maximum value, the node can increment the timeout interval by a delta value and can repeat the setting, the transmitting, the waiting, and the checking steps.

Abstract: Techniques for implementing Byzantine fault tolerance with verifiable secret sharing at constant overhead are provided. In one set of embodiments, a client can determine a secret value s to be shared with N replicas in a distributed system, s being input data for a service operation provided by the N replicas. The client can further encode s into an f-degree polynomial P(x) where f corresponds to a maximum number of faulty replicas in the distributed system, evaluate P(x) at i for i=1 to N resulting in N evaluations P(i), generate at least one f-degree recovery polynomial R(x) based on a distributed pseudo-random function (DPRF) f?(x), and evaluate R(x) at i for i=1 to N resulting in at least N evaluations R(i). The client can then invoke the service operation, the invoking comprising transmitting a message including P(i) and R(i) to each respective replica i.

Abstract: A shared log in a distributed system provides for direct access to the most current data state of an object stored in the shared log. Directly accessing the data state of an object obviates the need for a client to replay all the transactions stored in the shared log made on the object.

Abstract: In accordance with disclosed embodiments, a shared log system includes a sequencer that receives a source object and a snapshot time reference, where the source object is used to generate data for a destination object. The sequencer uses the snapshot time to determine whether the data state of the source object is current with respect to the snapshot time, to assess correctness of the generated data relative to the snapshot time.

Abstract: In accordance with disclosed embodiments, a shared log system includes a sequencer to verify transactions that comprise a source object and one or more members of the source object (source data objects), a target object and one or more members of the target object (target data objects), and a snapshot time reference. The sequencer verifies transaction using the snapshot time to determine whether the data states of the source data objects at the time of the snapshot time represent their most current data state in a shared log.

Abstract: The disclosure describes a failure-free execution agreement that includes n=3F+1 parties acting as replicas, and a number of parties acting as clients. One replica is designated as a primary. At most F replicas are presumed Byzantine faulty. The basic agreement protocol proceeds in three rounds: (1) client sends a request to the primary, who sends to all replicas; (2) each replica sends a threshold-part signature on hash to a first collector; (3) the collector combines the threshold-parts into a single signature and sends to all 3F+1 replicas which then commit and send to a second collector. The client proceeds when a signed block of requests arrives from the second collector.

Abstract: Techniques for implementing Byzantine fault tolerance with verifiable secret sharing at constant overhead are provided. In one set of embodiments, a client can determine a secret value s to be shared with N replicas in a distributed system, s being input data for a service operation provided by the N replicas. The client can further encode s into an f-degree polynomial P(x) where f corresponds to a maximum number of faulty replicas in the distributed system, evaluate P(x) at i for i=1 to N resulting in N evaluations P(i), generate at least one f-degree recovery polynomial R(x) based on a distributed pseudo-random function (DPRF) f?(x), and evaluate R(x) at i for i=1 to N resulting in at least N evaluations R(i). The client can then invoke the service operation, the invoking comprising transmitting a message including P(i) and R(i) to each respective replica i.

Abstract: Embodiments of the present disclosure relate to techniques for maintaining a state of a distributed system. In particular, certain embodiments relate to identifying a function. Some embodiments relate to, upon determining that the function comprises an annotation indicating that the function is capable of modifying the state of the distributed system, transforming the function to allow the function to generate updates to a state machine.