公开(公告)号：US20160077936A1

公开(公告)日：2016-03-17

申请号：US14484618

申请日：2014-09-12

Applicant: Facebook, Inc.

Inventor： Liyin Tang ,  Rishit Manit Shroff ,  Amitanand S. Aiyer ,  Arjen Roodselaar

IPC: G06F11/20 ,  G06F11/14

CPC classification number: G06F11/2033 ,  G06F11/1471 ,  G06F11/2028 ,  G06F11/2046 ,  G06F11/2097 ,  G06F2201/805 ,  G06F2201/84 ,  H04L67/1095 ,  H04L67/1097

Abstract: The disclosure is directed to failover mechanisms in a distributed computing system. A region of data is managed by multiple region servers. One of the region servers is elected as a “leader” and the remaining are “followers.” The leader serves the read/write requests from a client. The leader writes the data received from the client into the in-memory store and a local write-ahead log (“WAL”), and synchronously replicates the WAL to the followers. A region server designated as an “active” region server synchronizes a distributed data store with the data from the WAL. Active witness followers apply the data from the WAL to their in-memory store while shadow witness followers do not. Different types of servers provide failover mechanisms with different characteristics. A leader is elected based on their associated ranks—higher the rank, higher the likelihood of electing itself as a leader.

Abstract translation: 本公开涉及分布式计算系统中的故障切换机制。 数据区域由多个区域服务器管理。 其中一个区域服务器被选为“领导者”，其余的是“追随者”。领导者服务于客户端的读/写请求。 领导者将从客户端收到的数据写入内存中的存储和本地预写日志（“WAL”），同时将WAL复制到关注者。 指定为“活动”区域服务器的区域服务器将分布式数据存储与来自WAL的数据进行同步。 活跃的见证追随者将WAL中的数据应用到内存中，而影子见证追随者则不会。 不同类型的服务器提供具有不同特征的故障切换机制。 一个领导人是根据他们相关的职级选出的 - 排名越高，选举自己作为领导者的可能性就越高。

公开(公告)号：US09983957B2

公开(公告)日：2018-05-29

申请号：US15152769

申请日：2016-05-12

Applicant: Facebook, Inc.

Inventor： Liyin Tang ,  Rishit Manit Shroff ,  Amitanand S. Aiyer ,  Arjen Roodselaar

IPC: G06F11/20 ,  G06F11/14 ,  H04L29/08

CPC classification number: G06F11/2033 ,  G06F11/1471 ,  G06F11/2028 ,  G06F11/2046 ,  G06F11/2097 ,  G06F2201/805 ,  G06F2201/84 ,  H04L67/1095 ,  H04L67/1097

Abstract: The disclosure is directed to failover mechanisms in a distributed computing system. A region of data is managed by multiple region servers. One of the region servers is elected as a “leader” and the remaining are “followers.” The leader serves the read/write requests from a client. The leader writes the data received from the client into the in-memory store and a local write-ahead log (“WAL”), and synchronously replicates the WAL to the followers. A region server designated as an “active” region server synchronizes a distributed data store with the data from the WAL. Active witness followers apply the data from the WAL to their in-memory store while shadow witness followers do not. Different types of servers provide failover mechanisms with different characteristics. A leader is elected based on their associated ranks—higher the rank, higher the likelihood of electing itself as a leader.

公开(公告)号：US10521396B2

公开(公告)日：2019-12-31

申请号：US14996627

申请日：2016-01-15

Applicant: Facebook, Inc.

Inventor： Kannan Muthukkaruppan ,  Karthik Ranganathan ,  Liyin Tang

IPC: G06F16/10 ,  G06F16/13 ,  G06F16/178 ,  G06F16/182 ,  G06F16/22 ,  G06F16/185 ,  G06F16/174

Abstract: A region-based placement policy that can be used to achieve a better distribution of data in a clustered storage system is disclosed herein. The clustered storage system includes a master module to implement the region-based placement policy for storing one or more copies of a received data across many data nodes of the clustered storage system. When implementing the region-based placement policy, the master module splits the received data into one or more regions, where each region includes a contiguous portion of the received data. Further, for each of the plurality of regions, the master module stores complete copies of the region in a subset of the data nodes.

公开(公告)号：US20160253249A1

公开(公告)日：2016-09-01

申请号：US15152769

申请日：2016-05-12

Applicant: Facebook, Inc.

Inventor： Liyin Tang ,  Rishit Manit Shroff ,  Amitanand S. Aiyer ,  Arjen Roodselaar

IPC: G06F11/20 ,  H04L29/08

CPC classification number: G06F11/2033 ,  G06F11/1471 ,  G06F11/2028 ,  G06F11/2046 ,  G06F11/2097 ,  G06F2201/805 ,  G06F2201/84 ,  H04L67/1095 ,  H04L67/1097

Abstract: The disclosure is directed to failover mechanisms in a distributed computing system. A region of data is managed by multiple region servers. One of the region servers is elected as a “leader” and the remaining are “followers.” The leader serves the read/write requests from a client. The leader writes the data received from the client into the in-memory store and a local write-ahead log (“WAL”), and synchronously replicates the WAL to the followers. A region server designated as an “active” region server synchronizes a distributed data store with the data from the WAL. Active witness followers apply the data from the WAL to their in-memory store while shadow witness followers do not. Different types of servers provide failover mechanisms with different characteristics. A leader is elected based on their associated ranks—higher the rank, higher the likelihood of electing itself as a leader.

公开(公告)号：US09367410B2

公开(公告)日：2016-06-14

申请号：US14484618

申请日：2014-09-12

Applicant: Facebook, Inc.

Inventor： Liyin Tang ,  Rishit Manit Shroff ,  Amitanand S. Aiyer ,  Arjen Roodselaar

IPC: G06F11/07 ,  G06F11/20 ,  G06F11/14

CPC classification number: G06F11/2033 ,  G06F11/1471 ,  G06F11/2028 ,  G06F11/2046 ,  G06F11/2097 ,  G06F2201/805 ,  G06F2201/84 ,  H04L67/1095 ,  H04L67/1097

Abstract: The disclosure is directed to failover mechanisms in a distributed computing system. A region of data is managed by multiple region servers. One of the region servers is elected as a “leader” and the remaining are “followers.” The leader serves the read/write requests from a client. The leader writes the data received from the client into the in-memory store and a local write-ahead log (“WAL”), and synchronously replicates the WAL to the followers. A region server designated as an “active” region server synchronizes a distributed data store with the data from the WAL. Active witness followers apply the data from the WAL to their in-memory store while shadow witness followers do not. Different types of servers provide failover mechanisms with different characteristics. A leader is elected based on their associated ranks—higher the rank, higher the likelihood of electing itself as a leader.

Abstract translation: 本公开涉及分布式计算系统中的故障切换机制。 数据区域由多个区域服务器管理。 其中一个区域服务器被选为“领导者”，其余的是“追随者”。领导者服务于客户端的读/写请求。 领导者将从客户端收到的数据写入内存中的存储和本地预写日志（“WAL”），同时将WAL复制到关注者。 指定为“活动”区域服务器的区域服务器将分布式数据存储与来自WAL的数据进行同步。 活跃的见证追随者将WAL中的数据应用到内存中，而影子见证追随者则不会。 不同类型的服务器提供具有不同特征的故障切换机制。 一个领导人是根据他们相关的职级选出的 - 排名越高，选举自己作为领导者的可能性就越高。

公开(公告)号：US20160132518A1

公开(公告)日：2016-05-12

申请号：US14996627

申请日：2016-01-15

Applicant: Facebook, Inc.

Inventor： Kannan Muthukkaruppan ,  Karthik Ranganathan ,  Liyin Tang

IPC: G06F17/30

CPC classification number: G06F16/134 ,  G06F16/1752 ,  G06F16/178 ,  G06F16/183 ,  G06F16/185 ,  G06F16/22 ,  G06F16/2282

Abstract: A region-based placement policy that can be used to achieve a better distribution of data in a clustered storage system is disclosed herein. The clustered storage system includes a master module to implement the region-based placement policy for storing one or more copies of a received data across many data nodes of the clustered storage system. When implementing the region-based placement policy, the master module splits the received data into one or more regions, where each region includes a contiguous portion of the received data. Further, for each of the plurality of regions, the master module stores complete copies of the region in a subset of the data nodes.

Abstract translation: 这里公开了一种可用于在集群存储系统中实现更好的数据分发的基于区域的放置策略。 集群存储系统包括主模块，用于实现基于区域的放置策略，用于存储跨群集存储系统的许多数据节点的接收数据的一个或多个副本。 当实现基于区域的放置策略时，主模块将接收到的数据分割成一个或多个区域，其中每个区域包括所接收数据的连续部分。 此外，对于多个区域中的每一个，主模块在数据节点的子集中存储该区域的完整副本。

公开(公告)号：US09268808B2

公开(公告)日：2016-02-23

申请号：US13731722

申请日：2012-12-31

Applicant: Facebook, Inc.

Inventor： Kannan Muthukkaruppan ,  Karthik Ranganathan ,  Liyin Tang

IPC: G06F17/30

CPC classification number: G06F17/30094 ,  G06F17/30159 ,  G06F17/30174 ,  G06F17/30203 ,  G06F17/30221 ,  G06F17/30312 ,  G06F17/30339

Abstract: A region-based placement policy that can be used to achieve a better distribution of data in a clustered storage system is disclosed herein. The clustered storage system includes a master module to implement the region-based placement policy for storing one or more copies of a received data across many data nodes of the clustered storage system. When implementing the region-based placement policy, the master module splits the received data into one or more regions, where each region includes a contiguous portion of the received data. Further, for each of the plurality of regions, the master module stores complete copies of the region in a subset of the data nodes.

Abstract translation: 这里公开了一种可用于在集群存储系统中实现更好的数据分发的基于区域的放置策略。 集群存储系统包括主模块，用于实现基于区域的放置策略，用于存储跨群集存储系统的许多数据节点的接收数据的一个或多个副本。 当实现基于区域的放置策略时，主模块将接收到的数据分割成一个或多个区域，其中每个区域包括所接收数据的连续部分。 此外，对于多个区域中的每一个，主模块在数据节点的子集中存储该区域的完整副本。