Abstract: A method of reading data comprises receiving a request for a stripe of erasure coded data, sending read messages to at least a quorum of storage devices, and receiving at least the quorum of reply messages from the devices. The quorum of the reply messages includes at least a minimum number of stripe blocks needed to decode the data. The quorum meets a quorum condition of a number such that any two selections of the number of the stripe blocks intersect in the minimum number of the stripe blocks. A method of writing data comprises generating a timestamp, encoding the data, sending query messages including the timestamp to the storage devices, receiving query reply messages from each of at least the quorum of the devices, sending write messages to the devices, and receiving a write reply message from each of at least the quorum of the devices.

Abstract: A method of recovering a stripe of erasure coded data begins with sending query messages to storage devices. The method continues with receiving query reply messages from at least a first quorum of the storage devices. The query reply messages include a minimum number of the stripe blocks needed to decode the stripe. Following this, the stripe of erasure coded data is encoded. Next, a write message is sent to each of the storage devices, which include a timestamp and the stripe block destined for the storage device. The method concludes with receiving a write reply message from at least a second quorum of the storage devices indicating that the stripe block was successfully stored. The first and second quorums each meet a quorum condition of a number such that any two selections of the number of the stripe blocks intersect in the minimum number of the stripe blocks.

Abstract: A hierarchical method is provided for fault tolerance in a distributed computer system. A plurality of data centers is provided having a plurality of objects in each of the plurality of data centers. A local sub-protocol is used for dissemination of messages within a data center in the plurality of data centers and the local sub-protocol is activated from another data center of the plurality of data centers in a single round-trip message in the absence of faults.

Abstract: Data structure and timestamp management techniques for redundant storage. A plurality of storage devices are interconnected by a communication medium. At least two of the storage devices are designated devices for storing a block of data. Each designated device stores a version of the data and a first timestamp that is indicative of when the version of data was last updated. A second timestamp is indicative of a pending update to the block of data. When the update to the block of data is completed at one of the designated devices, the device discards the second timestamp. A storage device acting as coordinator instructs the device to discard the second timestamp. The designated storage devices store a plurality of blocks of data and corresponding timestamps according to a data structure. At least some of the entries in the data structure correspond to a range of data blocks that share a common timestamp. Entries in the data structure are arranged such that the ranges do not overlap.

Abstract: The present invention provides techniques for assignment and layout of redundant data in data storage system. In one aspect, the data storage system stores a number M of replicas of the data. Nodes that have sufficient resources available to accommodate a requirement of data to be assigned to the system are identified. When the number of nodes is greater than M, the data is assigned to M randomly selected nodes from among those identified. The data to be assigned may include a group of data segments and when the number of nodes is less than M, the group is divided to form a group of data segments having a reduced requirement. Nodes are then identified that have sufficient resources available to accommodate the reduced requirement. In other aspects, techniques are providing for adding a new storage device node to a data storage system having a plurality of existing storage device nodes and for removing data from a storage device node in such a data storage system.

Abstract: Read, write and recovery operations for replicated data are provided. In one aspect, a system for redundant storage of data included a plurality of storage devices and a communication medium for interconnecting the storage devices. At least two of the storage devices are designated devices for storing a block of data. Each designated device has a version of the data and a first timestamp that is indicative of when the version of data was last updated and a second timestamp that is indicative of any pending update to the block of data. The read, write and recovery operations are performed to the data using the first and second timestamps to coordinate the operations among the designated devices.

Abstract: Methods of processing transactions in a multi-tier network of at least one database client and a plurality of database servers including a primary and a backup database server are described. One method includes the step of concurrently executing a plurality of transactions including a selected transaction on a first database server during an epoch exclusively identifying the first database server as a primary. The selected transaction is committed on the first database server in response to a commit request, if the commit request is encountered before the epoch is changed. The selected transaction is broadcast to a second database server for non-concurrent execution. The selected transaction is committed on the second database server if execution is completed and the epoch has not changed.

Abstract: A method and system for providing exactly-once semantics for web-based transaction processing in a system that includes a client and a server. The client requests a form from the server. The server in response generates a unique identifier for identifying a particular transaction and provides a form with the unique identifier to the client. A user fills out the form and posts the filled out form to the server. Upon receiving the filled out form, the server generates a status page for informing the user that the transaction is being processed and returns the status page to the client. The status page includes client-side error handling logic. After returning the status page, the server performs transaction processing. In the event of failures, the error handling mechanism provides either exactly-one error handling semantics, which does not require user intervention, or at-most once error handling semantics that involves the user in error recovery.

Abstract: A computer system and software are provided for customizing on-line computer services. An application is able to interact with a sub-system between which an interceptor system can be interposed. The interceptor system includes a proxy interacting with the application, a customization module interacting with the proxy and a dispatcher interacting with the customization module and interacting with the sub-system. A customization repository contains the customization module and interacts with a service and a client sharing the application and the sub-system. A control interacts with the customization repository and the service for causing customization of the service and the client by disposing the proxy to interact with the application and the dispatcher to interact with the sub-system.

Abstract: A data service system includes web servers, each servicing any access request received by the data service system. Duplicate application servers are also provided, each processing any request directed from any one of the web servers. A session state information managing system is provided to allow different application servers to process requests belonging to a single session without requiring the requests to carry their entire session state information. The managing system includes a session state information manager that, when called by an application server in processing a request, (1) provides the session state information of the request to the application server, and (2) generate a state reference for a new session state information for the request after the application server has processed the request and generated the new session state information. The managing system also includes a store that stores all session state information.

Abstract: A distributed data center system protocol is provided which includes providing a client having a failure detector, a primary data center, and a backup data center operatively interconnected. A transaction operation is provided from the client to the primary database server in the primary data center and the backup database server for execution. If the client detects a failure or suspects a failure of the primary database server, the transaction operation is provided to the backup database server, which becomes the new primary database server. The database server executing the transaction operation returns the executed transaction operation to the client. If the primary data center suffers a disaster, the backup data center takes over control.

Abstract: A hierarchical method is provided for fault tolerance in a distributed computer system. A plurality of data centers is provided having a plurality of objects in each of the plurality of data centers. A local sub-protocol is used for dissemination of messages within a data center in the plurality of data centers and the local sub-protocol is activated from another data center of the plurality of data centers in a single round-trip message in the absence of faults.

Abstract: Methods of processing transactions in a multi-tier network of at least one database client and a plurality of database servers including a primary and a backup database server are described. One method includes the step of concurrently executing a plurality of transactions including a selected transaction on a first database server during an epoch exclusively identifying the first database server as a primary. The selected transaction is committed on the first database server in response to a commit request, if the commit request is encountered before the epoch is changed. The selected transaction is broadcast to a second database server for non-concurrent execution. The selected transaction is committed on the second database server if execution is completed and the epoch has not changed.

Abstract: A client method for providing three tier asynchronous client transparency includes sending a request and a count from a client to a first application server of an application server group. The request and count are resent to the group, if no corresponding decision is received within a pre-determined time period. The request is resent with an incremented count to a selected one of the first application server and the group, if the corresponding decision indicates that the result has not been committed. An application server method attempts to write a server identity to a write-once register in response to receipt of a client tier originated request. An attempt is made to write the computed result and a determined outcome as a first decision to the write-once register. The transaction identified by the request is terminated consistent with a second decision indicated by the write-once register. The result of the second decision is provided to the client-tier.

Abstract: A three-tiered transaction processing system is provided in which state information maintained in the database tier is used to facilitate recovery of transaction coordinators without the transaction coordinators having to maintain their own recovery information and a method of accessing such state information while controlling the lifetime of this recovery information; i.e., keeping it around long enough so that client tier will have a consistent view of the world before the information is forgotten. This solves the problem of outcome determination.

Abstract: A three-tier system is built up as: a single client application, a single server application, and a single database system. The server application provides a service that can be shared among multiple client applications. Server applications store their data in a database system. A client transparency mechanism and a server transparency mechanism are added. A database system on a clustered node is used for the database management. Server applications implement transactional behavior and the server side of the protocol so that the client applications may recover from server application and database system failures. The cluster application programming interface is used to determine when to retry. Information is stored in the database system so that the outcome of the transaction can be determined.

Abstract: A three-tier system is built up as: one or more client applications, one or more server applications, and one or more database systems. The server application provides a service that can be shared among multiple client applications. Server applications store their data in multiple databases. A client transparency mechanism and a server transparency mechanism are added. A database system on a clustered node is used for the database management. Server applications implement transactional behavior and the server side of the protocol so that the client applications may recover from server application and database system failures. A cluster application programming interface is used to determine when to retry. Information is stored in the server transparency mechanism so that the outcome of the transaction can be determined.

Abstract: A three-tier system is built up as: one or more client applications, one or more server applications, and one or more database systems. The server application provides a service that can be shared among multiple client applications. Server applications store their data in a database system. A client transparency mechanism and a server transparency mechanism are added. A database system on a clustered node is used for the database management. Server applications implement transactional behavior and the server side of the protocol so that the client applications may recover from server application and database system failures. The cluster application programming interface is used to determine whether, when, and if to retry. Information is stored in the database system so that the outcome of the transaction can be determined.