公开(公告)号：US09189510B2

公开(公告)日：2015-11-17

申请号：US13777814

申请日：2013-02-26

Applicant: Facebook, Inc.

Inventor： Yee Jiun Song ,  Philippe Ajoux ,  Harry C. Li ,  Jason Sobel ,  Sanjeev Kumar ,  Rajesh Nishtala

IPC: G06F17/30

CPC classification number: G06F12/0802 ,  G06F3/0619 ,  G06F3/065 ,  G06F3/067 ,  G06F17/30345 ,  G06F17/30575 ,  G06F17/30876 ,  G06F2212/1032

Abstract: When multiple regional data clusters are used to store data in a system, maintaining cache consistency across different regions is important for providing a desirable user experience. In one embodiment, there is a master data cluster where all data writes are performed, and the writes are replicated to each of the slave data clusters in the other regions. Appended to the replication statements are invalidations for cache values for the keys whose values have been changed in the master data cluster. An apparatus in the master data cluster logs replication statements sent to the slave databases. When a slave database fails, the apparatus extracts the invalidations intended for the failed database and publishes the invalidations to a subscriber in the region of the failed database. The subscriber sends the invalidations to the local caches to cause stale data for those keys to be deleted from the caches.

Abstract translation: 当多个区域数据集群用于在系统中存储数据时，在不同区域之间保持高速缓存的一致性对于提供理想的用户体验是重要的。 在一个实施例中，存在执行所有数据写入的主数据集群，并且将写入复制到其他区域中的每个从属数据集群。 附加到复制语句对于其值在主数据集群中已更改的键的缓存值无效。 主数据集群中的一个装置记录发送到从属数据库的复制语句。 当从属数据库发生故障时，设备将提取针对故障数据库的无效，并在失败的数据库区域中向用户发布无效。 用户将无效发送到本地缓存，以使这些密钥的过期数据从高速缓存中删除。

公开(公告)号：US20150156158A1

公开(公告)日：2015-06-04

申请号：US14620085

申请日：2015-02-11

Applicant: Facebook, Inc.

Inventor： Harry C. Li ,  Philippe Vincent Ajoux ,  Laurent Demailly ,  Liat Atsmon Guz ,  Yee Jiun Song ,  Sabyasachi Roy ,  Sanjeev Kumar

IPC: H04L12/58 ,  H04L29/08

CPC classification number: H04L67/26 ,  G06F17/30 ,  H04L51/00 ,  H04L51/26 ,  H04L51/32 ,  H04L67/00 ,  H04L67/22 ,  H04L67/306 ,  H04L69/16 ,  H04W4/08

Abstract: Disclosed is a technology for publishing subscriptions in a publish-subscribe system in accordance with subscription groups. The technology may include (i) determining a consumption characteristic by which each of multiple subscribers in a publish-subscribe system consumes a subscription published by a publisher; (ii) identifying the subscribers whose consumption characteristics satisfy a specified similarity criterion; (iii) defining multiple subscription groups, each of which includes subscriptions of those of the subscribers whose consumption characteristics satisfy the specified similarity criterion; (iv) storing the subscriptions in multiple logical partitions of a storage system where each of the logical partitions contains a separate non-overlapping subset of the subscriptions; and (v) transmitting the subscriptions to the subscribers in accordance with the subscription groups.

Abstract translation: 公开了一种根据订阅组在发布订阅系统中发布订阅的技术。 该技术可以包括（i）确定消费特征，通过该消费特征，发布订阅系统中的多个订阅者中的每一个消费出版者发布的订阅; （ii）识别其消费特征满足指定的相似性标准的用户; （iii）定义多个订阅组，每个订阅组包括对消费特征满足指定的相似性标准的用户的订阅; （iv）将订阅存储在存储系统的多个逻辑分区中，其中每个逻辑分区包含订阅的单独的非重叠子集; 和（v）根据订阅团体向用户发送订阅。

公开(公告)号：US20150180719A1

公开(公告)日：2015-06-25

申请号：US14450148

申请日：2014-08-01

Applicant: Facebook, Inc.

Inventor： Qiang Wu ,  Sanjeev Kumar ,  Sachin Kadloor

IPC: H04L12/24 ,  H04L12/911 ,  H04L12/26

CPC classification number: H04L67/1031 ,  G06F9/5083 ,  H04L41/0896 ,  H04L43/08 ,  H04L43/0817 ,  H04L47/726

Abstract: A self-adaptive control system based on proportional-integral (PI) control theory for dynamic capacity management of latency-sensitive application servers (e.g., application servers associated with a social networking application) are disclosed. A centralized controller of the system can adapt to changes in request rates, changes in application and/or system behaviors, underlying hardware upgrades, etc., by scaling the capacity of a cluster up or down so that just the right amount of capacity is maintained at any time. The centralized controller uses information relating to a current state of the cluster and historical information relating to past state of the cluster to predict a future state of the cluster and use that prediction to determine whether to scale up or scale down the current capacity to reduce latency and maximize energy savings. A load balancing system can then distribute traffic among the servers in the cluster using any load balancing methods.

Abstract translation: 公开了一种基于比例积分（PI）控制理论的自适应控制系统，用于对等待时间敏感的应用服务器（例如，与社交网络应用相关联的应用服务器）的动态容量管理。 系统的集中式控制器可以通过将集群的容量上下扩展来适应请求速率的变化，应用和/或系统行为的变化，底层的硬件升级等，从而保持正确的容量。 随时。 集中控制器使用与集群的当前状态有关的信息和与集群的过去状态相关的历史信息来预测集群的未来状态，并使用该预测来确定是放大还是缩小当前容量以减少等待时间 并最大限度地节省能源。 然后，负载平衡系统可以使用任何负载平衡方法在群集中的服务器之间分配流量。

公开(公告)号：US20140244584A1

公开(公告)日：2014-08-28

申请号：US13777814

申请日：2013-02-26

Applicant: Facebook, Inc.

Inventor： Yee Jiun Song ,  Philippe Ajoux ,  Harry C. Li ,  Jason Sobel ,  Sanjeev Kumar ,  Rajesh Nishtala

IPC: G06F17/30

CPC classification number: G06F12/0802 ,  G06F3/0619 ,  G06F3/065 ,  G06F3/067 ,  G06F17/30345 ,  G06F17/30575 ,  G06F17/30876 ,  G06F2212/1032

Abstract: When multiple regional data clusters are used to store data in a system, maintaining cache consistency across different regions is important for providing a desirable user experience. In one embodiment, there is a master data cluster where all data writes are performed, and the writes are replicated to each of the slave data clusters in the other regions. Appended to the replication statements are invalidations for cache values for the keys whose values have been changed in the master data cluster. An apparatus in the master data cluster logs replication statements sent to the slave databases. When a slave database fails, the apparatus extracts the invalidations intended for the failed database and publishes the invalidations to a subscriber in the region of the failed database. The subscriber sends the invalidations to the local caches to cause stale data for those keys to be deleted from the caches.

Abstract translation: 当多个区域数据集群用于在系统中存储数据时，在不同区域之间保持高速缓存的一致性对于提供理想的用户体验是重要的。 在一个实施例中，存在执行所有数据写入的主数据集群，并且将写入复制到其他区域中的每个从属数据集群。 附加到复制语句对于其值在主数据集群中已更改的键的缓存值无效。 主数据集群中的一个装置记录发送到从属数据库的复制语句。 当从属数据库发生故障时，设备将提取针对故障数据库的无效，并在失败的数据库区域中向用户发布无效。 用户将无效发送到本地缓存，以使这些密钥的过期数据从高速缓存中删除。