简体繁体 English

Hbase是否为每个column-family或columnFamily：Column创建HFile？

[英]Does Hbase create a HFile for each column-family or columnFamily:Column?

原文 2019-04-02 17:39:34 6 1 hbase/ column-family/ hfile

I am trying to understand the Hbase architecture with respect to Logical data model vs Physical data storage. 我试图了解有关逻辑数据模型与物理数据存储的Hbase体系结构。 I am little confused about the HFile creation. 我对HFile的创建感到困惑。 If we have a column family with 2 columns, does Hbase create 2 HFiles or just one? 如果我们有一个包含2列的列族，那么Hbase是创建2个HFile还是仅创建1个？

Below is the diagram that I was looking and example below shows logical to physical mapping for each cf:col. 下面是我一直在查看的图，下面的示例显示了每个cf：col的逻辑到物理映射。 Please help me clear this confusion 请帮助我清除这种混乱

https://mapr.com/blog/hbase-and-mapr-db-designed-distribution-scale-and-speed/assets/blogimages/Logical-vs-physical-storage.png https://mapr.com/blog/hbase-and-mapr-db-designed-distribution-scale-and-speed/assets/blogimages/Logical-vs-physical-storage.png

1 个解决方案

HFiles are created on a column-family basis, so cf1:a and cf1:b would be stored in the same HFile if they are in the same region, but cf2:a would be in a different one. HFile是基于列族创建的，因此，如果cf1:a和cf1:b位于同一区域中，则它们将存储在同一HFile中，但是cf2:a将位于不同的区域中。

On the diagram Address:street and Address:city are both part of the Address column family and the data will end up in the same HFile. 在图上， Address:street和Address:city都是Address列族的一部分，并且数据将最终存储在同一HFile中。

This is the same for the MemStore, so for each column family on a single RegionServer there will be a separate MemStore instance. 这与MemStore相同，因此对于单个RegionServer上的每个列族，将有一个单独的MemStore实例。