公开(公告)号：US09146868B1

公开(公告)日：2015-09-29

申请号：US13743800

申请日：2013-01-17

申请人： Symantec Corporation

发明人： Mithlesh Thukral ,  Mukesh Bafna

IPC分类号： G06F12/00 ,  G06F13/00 ,  G06F13/28 ,  G06F12/08

CPC分类号： G06F12/0804 ,  G06F12/0866 ,  G06F2212/1032 ,  G06F2212/314

摘要： A computer-implemented method for eliminating inconsistencies between backing stores and caches may include (1) detecting at least one inconsistency between a backing store and a cache, (2) identifying a synchronization marker that bifurcates write operations queued in the cache into (i) a subset of one or more write operations known to have been successfully performed on the backing store and (ii) an additional subset of one or more additional write operations not yet known to have been successfully performed on the backing store, (3) identifying the additional subset of additional write operations based at least in part on the synchronization marker, (4) performing the additional subset of additional write operations on the backing store, and then (5) updating the synchronization marker based at least in part on performing the additional subset of additional write operations on the backing store. Various other methods, systems, and computer-readable media are also disclosed.

摘要翻译： 用于消除后备存储和高速缓存之间的不一致性的计算机实现的方法可以包括（1）检测后备存储和高速缓存之间的至少一个不一致性，（2）识别将高速缓存中排队的写入操作分叉为（i） 已知在后备存储上已经成功执行的一个或多个写入操作的子集，以及（ii）尚未已知在后备存储器上成功执行的一个或多个附加写操作的附加子集，（3）识别 至少部分地基于同步标记的附加写入操作的附加子集，（4）在后备存储上执行附加写入操作的附加子集，然后（5）至少部分地基于执行附加写入操作来更新同步标记 后备存储上附加写入操作的子集。 还公开了各种其它方法，系统和计算机可读介质。

公开(公告)号：US09378141B1

公开(公告)日：2016-06-28

申请号：US13857496

申请日：2013-04-05

申请人： Symantec Corporation

发明人： Mithlesh Thukral ,  Mukesh Bafna ,  Shirish Vijayvargiya ,  Sanjay Jain ,  Sushil Patil ,  Sanjay Kumar ,  Anindya Banerjee

IPC分类号： G06F12/08 ,  H04L29/08

CPC分类号： G06F12/0813 ,  G06F12/0862 ,  G06F2212/1024 ,  G06F2212/314 ,  G06F2212/466 ,  H04L67/2842 ,  H04L67/2847

摘要： Caching metadata that identify hot blocks at a per local cache level are tracked. Tracked caching metadata are maintained so as to be persistent and shared across nodes of the cluster. Local caches are pre-warmed by using maintained caching metadata, responsive to detecting specific node level events. Such events can result in hot blocks being absent from a local cache, such as a failover between nodes or an unexpected failure local to a specific node. Another event example is the access of shared storage content, such as opening a file or mounting a file system by a specific node, in response to which the associated local cache can be pre-warmed using the tracked caching metadata for the specific file, or for each file of the file system. To pre-warm a local cache, hot blocks of stored content identified by corresponding caching metadata are loaded into the local cache.

摘要翻译： 跟踪在每个本地缓存级别标识热块的缓存元数据。 维护跟踪的缓存元数据，以便在集群的节点之间持久化和共享。 响应于检测到特定节点级事件，本地缓存通过使用维护的缓存元数据进行预热。 这种事件可导致本地缓存中不存在热块，例如节点之间的故障转移或特定节点本地的意外故障。 另一个事件示例是共享存储内容的访问，例如打开文件或由特定节点装载文件系统，响应于该特定节点可以使用跟踪的缓存元数据预先对相关联的本地高速缓存进行预热，或者 对于文件系统的每个文件。 要预热本地缓存，由相应的缓存元数据标识的存储内容的热块将加载到本地缓存中。