Abstract: Throughput is preserved in a distributed system while maintaining concurrency by pushing a commit wait period to client commit paths and to future readers. As opposed to servers performing commit waits, the servers assign timestamps, which are used to ensure that causality is preserved. When a server executes a transaction that writes data to a distributed database, the server acquires a user-level lock, and assigns the transaction a timestamp equal to a current time plus an interval corresponding to bounds of uncertainty of clocks in the distributed system. After assigning the timestamp, the server releases the user-level lock. Any client devices, before performing a read of the written data, must wait until the assigned timestamp is in the past.

Abstract: The present technology proposes techniques for generating globally coherent timestamps. This technology may allow distributed systems to causally order transactions without incurring various types of communication delays inherent in explicit synchronization. By globally deploying a number of time masters that are based on various types of time references, the time masters may serve as primary time references. Through an interactive interface, the techniques may track, calculate and record data relative to each time master thus providing the distributed systems with causal timestamps.

Abstract: A system, computer-readable storage medium storing at least one program, and a computer-implemented method for identifying a storage group in a distributed storage system into which data is to be stored is presented. A data structure including information relating to storage groups in a distributed storage system is maintained, where a respective entry in the data structure for a respective storage group includes placement metrics for the respective storage group. A request to identify a storage group into which data is to be stored is received from a computer system. The data structure is used to determine an identifier for a storage group whose placement metrics satisfy a selection criterion. The identifier for the storage group whose placement metrics satisfy the selection criterion is returned to the computer system.

Abstract: A method reads and writes data from a database table. Each row in the table has a primary key and multiple non-key columns. Each non-key column has one or more column values, and each column value has an associated timestamp that identifies when the column value was stored. The timestamps associated with the column values in each non-key column provide a unique order for the column values. A read transaction is initiated to read from a first non-key column of a first row. A write transaction is in progress that is updating a second non-key column of the first row, where the second non-key column is distinct from the first non-key column. The write transaction holds a lock on the second non-key column of the first row. The method concurrently reads the data from the first non-key column and writes a new column value to the second non-key column.

Abstract: The present technology proposes techniques for ensuring globally consistent transactions. This technology may allow distributed systems to ensure the causal order of read and write transactions across different partitions of a distributed database. By assigning causally generated timestamps to the transactions based on one or more globally coherent time services, the timestamps can be used to preserve and represent the causal order of the transactions in the distributed system. In this regard, certain transactions may wait for a period of time after choosing a timestamp in order to delay the start of any second transaction that might depend on it. The wait may ensure that the effects of the first transaction are not made visible until its timestamp is guaranteed to be in the past. This may ensure that a consistent snapshot of the distributed database can be determined for any past timestamp.

Abstract: Paxos transactions are pipelined in a distributed database formed by a plurality of replica servers. A leader server is selected by consensus of the replicas, and receives a lock on leadership for an epoch. The leader gets Paxos log numbers for the current epoch, which are greater than the numbers allocated in previous epochs. The leader receives database write requests, and assigns a Paxos number to each request. The leader constructs a proposed transaction for each request, which includes the assigned Paxos number and incorporates the request. The leader transmits the proposed transactions to the replicas. Two or more write requests that access distinct objects in the database can proceed simultaneously. The leader commits a proposed transaction to the database after receiving a plurality of confirmations for the proposed transaction from the replicas. After all the Paxos numbers have been assigned, inter-epoch tasks are performed before beginning a subsequent epoch.

Abstract: A system, computer-readable storage medium storing at least one program, and a computer-implemented method for performing operations on target servers is presented. A request including an operation is received. A set of target servers associated with the operation is identified. The following request processing operations are performed until a predetermined termination condition has been satisfied: a target server in the set of target servers to which the request has not been issued and whose health metrics satisfy health criteria is identified, the request to perform the operation is issued to the target server, and when the request to perform the operation fails at the target server, health metrics for the target server are updated to indicate that the request to perform the operation failed at the target server and health check operation is scheduled to be performed with respect to the target server.

Abstract: Paxos transactions are pipelined in a distributed database formed by a plurality of replica servers. A leader server is selected by consensus of the replicas, and receives a lock on leadership for an epoch. The leader gets Paxos log numbers for the current epoch, which are greater than the numbers allocated in previous epochs. The leader receives database write requests, and assigns a Paxos number to each request. The leader constructs a proposed transaction for each request, which includes the assigned Paxos number and incorporates the request. The leader transmits the proposed transactions to the replicas. Two or more write requests that access distinct objects in the database can proceed simultaneously. The leader commits a proposed transaction to the database after receiving a plurality of confirmations for the proposed transaction from the replicas. After all the Paxos numbers have been assigned, inter-epoch tasks are performed before beginning a subsequent epoch.

Abstract: The present technology proposes techniques for generating globally coherent timestamps. This technology may allow distributed systems to causally order transactions without incurring various types of communication delays inherent in explicit synchronization. By globally deploying a number of time masters that are based on various types of time references, the time masters may serve as primary time references. Through an interactive interface, the techniques may track, calculate and record data relative to each time master thus providing the distributed systems with causal timestamps.

Abstract: A distributed storage system is provided. The distributed storage system includes multiple front-end servers and zones for managing data for clients. Data within the distributed storage system is associated with a plurality of accounts and divided into a plurality of groups, each group including a plurality of splits, each split being associated with a respective account, and each group having multiple tablets and each tablet managed by a respective tablet server of the distributed storage system. Data associated with different accounts may be replicated within the distributed storage system using different data replication policies. There is no limit to the amount of data for an account by adding new splits to the distributed storage system. In response to a client request for a particular account's data, a front-end server communicates such request to a particular zone that has the client-requested data and returns the client-requested data to the requesting client.

Abstract: The present technology proposes techniques for generating globally coherent timestamps. This technology may allow distributed systems to causally order transactions without incurring various types of communication delays inherent in explicit synchronization. By globally deploying a number of time masters that are based on various types of time references, the time masters may serve as primary time references. Through an interactive interface, the techniques may track, calculate and record data relative to each time master thus providing the distributed systems with causal timestamps.

Abstract: A method reads and writes data from a database table. Each row in the table has a primary key and multiple non-key columns. Each non-key column has one or more column values, and each column value has an associated timestamp that identifies when the column value was stored. The timestamps associated with the column values in each non-key column provide a unique order for the column values. A read transaction is initiated to read from a first non-key column of a first row. A write transaction is in progress that is updating a second non-key column of the first row, where the second non-key column is distinct from the first non-key column. The write transaction holds a lock on the second non-key column of the first row. The method concurrently reads the data from the first non-key column and writes a new column value to the second non-key column.

Abstract: A system, computer-readable storage medium storing at least one program, and a computer-implemented method for identifying a storage group in a distributed storage system into which data is to be stored is presented. A data structure including information relating to storage groups in a distributed storage system is maintained, where a respective entry in the data structure for a respective storage group includes placement metrics for the respective storage group. A request to identify a storage group into which data is to be stored is received from a computer system. The data structure is used to determine an identifier for a storage group whose placement metrics satisfy a selection criterion. The identifier for the storage group whose placement metrics satisfy the selection criterion is returned to the computer system.

Abstract: A distributed storage system is provided. The distributed storage system includes multiple front-end servers and zones for managing data for clients. Data within the distributed storage system is associated with a plurality of accounts and divided into a plurality of groups, each group including a plurality of splits, each split being associated with a respective account, and each group having multiple tablets and each tablet managed by a respective tablet server of the distributed storage system. Data associated with different accounts may be replicated within the distributed storage system using different data replication policies. There is no limit to the amount of data for an account by adding new splits to the distributed storage system. In response to a client request for a particular account's data, a front-end server communicates such request to a particular zone that has the client-requested data and returns the client-requested data to the requesting client.