Google Inc.

Vishal Kasera

Amit Agarwal

Jean-Michel Leon

Kenneth M Ashcraft

Uri N Lerner

Michael Jahr

A database server receives a request from a client application for performing a data transaction on persistent data storage. The request is sent to a set of replication servers. An acknowledgement for the request is received from each replication server, including a start sequence number and an end sequence number for data that is stored in local cache of the replication server, and a latest committed sequence number for data that was written to the persistent data storage by the replication server. A maximum value of latest committed sequence numbers received from the set of replication servers is determined. For each replication server, it is examined whether there is a gap between the start sequence number for data stored in local cache and the maximum value of the latest committed sequence numbers. Based on the examining, it is determined whether there is an occurrence of loss of data.

Database replication

A write request is received at a database server from a client application for writing data to persistent data storage. In response to receiving the write request, the database server selects a set of multiple replication servers. The data is sent from the database server to the selected set of multiple replication servers for writing to the persistent data storage. Confirmation is received at the database server from replication servers in the selected set of multiple replication servers. In response to receiving confirmation from the replication servers in the selected set of multiple replication servers, the database server sends to the client application information indicating success of the write request.

A replication server receives a first write request from a database server for writing first data to persistent data storage. The replication server determines that the first write request is designated as asynchronous. In response to determining that the first write request is asynchronous, the first data is saved to local cache at the replication server and an acknowledgement is sent to the database server. The local cache stores second data to be written to persistent data storage when the first data is saved to the local cache. The first data and the second data are written from the local cache to the persistent data storage in a batch operation.

One embodiment of the present invention provides a system that selectively deletes clusters of conceptually-related words from a probabilistic generative model for textual documents. During operation, the system receives a current model, which contains terminal nodes representing random variables for words and contains one or more cluster nodes representing clusters of conceptually related words. Nodes in the current model are coupled together by weighted links, so that if an incoming link from a node that has fired causes a cluster node to fire with a probability proportionate to a weight of the incoming link, an outgoing link from the cluster node to another node causes the other node to fire with a probability proportionate to the weight of the outgoing link. Next, the system processes a given cluster node in the current model for possible deletion. This involves determining a number of outgoing links from the given cluster node to terminal nodes or cluster nodes in the current model. If the determined number of outgoing links is less than a minimum value, or if the frequency with which the given cluster node fires is less than a minimum frequency, the system deletes the given cluster node from the current model.

Growing community of inventors

Mountain View, CA, United States of America

Vishal Kasera