英文:
RocksDBStateBackend supports incremental checkpoint while FsStateBackend doesn't
问题
对于FsStateBackend，本地检查点数据保存在内存中，而远程持久化检查点数据保存在分布式文件系统中，比如HDFS。
对于RocksDBStateBackend，本地检查点数据保存在本地的RocksDB中，而远程持久化检查点数据也保存在分布式文件系统中，比如HDFS。
所以，我想知道为什么FsStateBackend不提供增量检查点功能，因为它与RocksDBStateBackend相同，它们都将远程持久化检查点数据存储在分布式文件系统上，比如HDFS。
英文:
For FsStateBackend, local checkpoint data are saved in memory, and remote persistent checkpoint data are saved in distributed file system,such as HDFS
For RocksDBStateBackend,local checkpoint data are saved in local RocksDB, and remote persistent checkpoint data is also saved in distributed file system,such as HDFS.
So, I wonder why FsStateBackend doesn't provide the incremental checkpoint capabilities, since it is the same as RocksDBStateBackend,both of them store remote persistent checkpoint data on the distributed file system like HDFS.
答案1
得分: 1
FsStateBackend 是一个传统类，等同于使用 HashMapStateBackend 和 FileSystemCheckpointStorage（请参阅此文档）。增量检查点的问题在于 HashMapStateBackend 尚不支持（但正在进行工作以添加支持）。
英文:
FsStateBackend is a legacy class, equivalent to using HashMapStateBackend and FileSystemCheckpointStorage (see this documentation). The issue with incremental checkpoints is that it's not (yet) supported by the HashMapStateBackend, though work is underway to add that.
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论