Google Inc.

Alexander Kesselman

Alexandre Drobychev

George Datuashvili

Frank Dachille

Rebekah C Vickrey

Yonatan Zunger

Gia Datuashvili

Michael O'Reilly

Dmitry Broyde

Zia Syed

Angus Lees

Zunger Yonatan

Daniel J Ford

Arun Subbiah

Stefan V Gheorghita

Frank C Dachille

Michael O'REILLY

A method is performed by a device of a group of devices in a distributed data replication system. The method includes storing an index of objects in the distributed data replication system, the index being replicated while the objects are stored locally by the plurality of devices in the distributed data replication system. The method also includes conducting a scan of at least a portion of the index and identifying a redundant replica(s) of the at least one of the objects based on the scan of the index. The method further includes de-duplicating the redundant replica(s), and updating the index to reflect the status of the redundant replica.

Asynchronous distributed de-duplication for replicated content addressable storage clusters

A system, computer-readable storage medium storing at least one program, and a computer-implemented method for facilitating deduplication of operations to be performed is presented. An operation to be performed is received. A mapping function is applied to at least one parameter of the operation to produce a mapping value in a target mapping space, the target mapping space being partitioned between target servers in a set of target servers proportional to resource capacities of the target servers in the set of target servers. A target server in the set of target servers whose portion of the target mapping space includes the mapping value is identified. The operation is issued to the target server.

System and method for facilitating deduplication of operations to be performed

The various embodiments described herein include methods, systems, and devices for managing load in a distributed storage system. In one aspect, a method is performed at a first instance server in the distributed storage system, the first instance server having memory and at least one processor coupled to the memory. The method includes: (i) issuing a first plurality of requests to a second instance server; (ii) obtaining one or more messages from the second instance server in response to the first plurality of requests, the messages indicating a utilization rate of the second instance server; (iii) determining a transaction rate limit for the second instance server based on the utilization rate of the second instance server; and (iv) issuing a second plurality of requests to the second instance server, where the second plurality of requests are issued at a rate no greater than the transaction rate limit.

System and method for managing load in a distributed storage system

A distributed storage system has multiple instances. There is a plurality of local instances, and at least some of the local instances are at physically distinct geographic locations. Each local instance is configured to store data for a non-empty set of blobs in a plurality of data stores having a plurality of distinct data store types. In addition, each local instance stores metadata for the respective set of blobs in a metadata store distinct from the data stores. There is also a plurality of global instances. Each global instance is configured to store data for zero or more blobs in zero or more data stores and store metadata for all blobs stored at any local or global instance. The system selects one global instance to run a replication module that replicates blobs between instances according to blob policies. Some systems also include dynamic replication based on user needs.

Storage of data in a distributed storage system

A distributed storage system has a plurality of instances. A computer system simulates the state of the distributed storage system. The system obtains a current state of the distributed storage system and replication policies for objects in the distributed storage system. Each replication policy specifies criteria for placing copies of the relevant objects among the plurality of instances. The system receives proposed modifications to the state of the distributed storage system and simulates the state of the distributed storage system over time based on the current state of the distributed storage system, current statistical trends in the state of the distributed storage system, the replication policies for the objects in the distributed storage system, and the proposed modifications to the state of the distributed storage system. One or more reports are generated relating to time evolution of the state of the distributed storage system based on the simulation.

Systems and methods of simulating the state of a distributed storage system

Managing placement of object replicas is performed at a first instance of a distributed storage system. One or more journals are opened for storage of object chunks. Each journal is associated with a single placement policy. A first object is received comprising at least a first object chunk. The first object is associated with a first placement policy. The first object chunk is stored in a first journal whose associated placement policy matches the first placement policy. The first journal stores only object chunks for objects whose placement policies match the first placement policy. For the first journal, the receiving and storing operations are repeated for multiple objects whose associated placement policies match the first placement policy, until a first termination condition occurs. Then, the first journal is closed. Subsequently, the first journal is replicated to a second instance of the distributed storage system according to the first placement policy.

Grouping of objects in a distributed storage system based on journals and placement policies

Placement of object replicas in a distributed storage system includes, at a first instance, opening a journal for storage of object chunks. An object is received, which comprises one or more chunks. Each chunk comprises one or more storage blocks. The blocks for a single chunk are stored in a single journal. Global metadata for the object is stored, which includes a list of chunks for the object. Local metadata for the chunk is stored, which includes a block list identifying each block of the plurality of blocks. The local metadata is associated with the journal. The journal is later closed. The journal is subsequently replicated to a second instance. The global metadata is updated to reflect the replication, whereas the local metadata is unchanged by the replication.

Hierarchical chunking of objects in a distributed storage system

Managing consistency of object replicas is performed at a first instance of a distributed storage system. The first instance performs garbage collection on a shard that includes a first plurality of object chunks, thereby removing a second plurality of object chunks from the shard. This leaves a third plurality of object chunks in the shard, where the first plurality of object chunks is the union of the second and third pluralities of object chunks. The first instance sends a first list of identifiers to a second instance of the distributed storage system. The second instance has a replica of the shard. The first list of identifiers specifies the object chunks in the third plurality of object chunks. The second instance removes all object chunks from the replica of the shard that are not included in the first list.

Weakly synchronized garbage collection and compaction for aggregated, replicated object stores

A system, computer-readable storage medium storing at least one program, and a computer-implemented method for managing load on a downstream server in a distributed storage system is presented. Messages received from the downstream server in response to transactions issued by the upstream server to the downstream server are monitored, where the transactions are issued to the downstream server at a rate no greater than a transaction rate limit, the transaction rate limit being based at least in part on a maximum transaction processing rate of the downstream server. The transaction rate limit is adjusted based at least in part on a number of messages that include a pushback signal received over at least one predetermined time interval, where the pushback signal indicates that a utilization rate of the downstream server has exceeded a predetermined utilization rate.

System and method for managing load on a downstream server in a distributed storage system

A system, computer-readable storage medium storing at least one program, and a computer-implemented method for storing metadata for a file in a distributed storage system is presented. Metadata for a file is generated. A subset of metadata clusters in a distributed storage system is identified to form a quorum of metadata clusters. Requests to store the metadata for the file on the metadata clusters in the distributed storage system are issued, where the requests include high priority requests to store the metadata for the file on at least one of the metadata clusters in the quorum of metadata clusters and normal priority requests to store the metadata in the remaining metadata clusters, the metadata being stored on the at least one of the metadata clusters in the quorum of metadata clusters before being stored on the remaining metadata clusters.

System and method for storing metadata for a file in a distributed storage system

A system and method for generating replication requests for objects in a distributed storage system is provided. For a respective object in a distributed storage system the following is performed. Replication policies for the object that have not been satisfied are determined. Replication requests are ranked for the object whose replication policies have not been satisfied based on a number of replicas of the object that need to be created in order to satisfy the replication policies for the object. Replication requests are generated for the object based at least in part on the replication policies for the object that have not been satisfied and on a current state of the distributed storage system. At least a subset of the replication requests for the objects in the distributed storage system are distributed to respective instances of the distributed storage system corresponding to the replication requests for execution.

Location assignment daemon (LAD) for a distributed storage system

A method for uploading an object into a distributed storage system is implemented at a computing device The computing device splits an object into one or more chunks and uploads the one or more chunks into the distributed storage system. For each uploaded chunk, the computing device receives a write token from the distributed storage system, inserts an entry into an extents table of the object for the chunk in accordance with the received write token and the chunk ID, chunk offset, and chunk size of the chunk, generates a digest of the extents table, the digest representing the one or more chunks that the client expects to be within the distributed storage system, and sends the digest of the extents table to the distributed storage system. The distributed storage system is configured to use the digest to determine whether it has each of the one or more client-expected chunks.

Method and system for uploading data into a distributed storage system

A method allocates object replicas in a distributed storage system. The method identifies a plurality of objects in the distributed storage system. Each object has an associated storage policy that specifies a target number of object replicas stored at distinct instances of the distributed storage system. The method identifies an object of the plurality of objects whose number of object replicas exceeds the target number of object replicas specified by the storage policy associated with the object. The method selects a first replica of the object for removal based on last access times for replicas of the object, and transmits a request to a first instance of the distributed storage system that stores the first replica. The request instructs the first instance to remove the first replica of the object.

Growing community of inventors

Sunnyvale, CA, United States of America

Alexander Kesselman