Transfer: https://www.cnblogs.com/hbase-community/p/8629222.html
HBase Overview
HBase is a distributed storage, database engine can support millions of QPS, PB levels of storage, these have been validated in a production environment, and in the majority of the company has been verified. Especially Ali, millet, Jingdong, interior, thousands of drops, HBase cluster on a million units. Select a technology first condition is aligned big companies, big companies will invest a lot of manpower to maintain, improve and contribute to the community.
NewSQL on the relationship with the NoSQL
Technology is always evolving forward, now in NewSQL mention, in fact, in my opinion is a NewSQL package over NoSQL, a sub-scene. NoSQL in a large table, is typically provided KV1V2 ...... Vn, where each V may be. IB, may be 100MB. It can be said is a meta-existence, 01 is similar to the digital world, in any combination. In order HBase represented in NoSQL, HBase can be combined in any scene, NewSQL may be added on top of SQL layer or a layer closer to the scene to add a child transaction.
Separation on computing and storage
On the cloud, for most engine core is stored in separate computing, storage on demand billing, at least have elastic stretch. Computing node stores press provided. If full accordance with QPS billing or costs were prohibitively high, or more difficult to meet the scene, such as storage 10M, in the end counted once QPS, or how many times. Since HBase storage computing is inherently separate, comparative framework on the natural adaptation cloud, we can say that the cloud, HBase has more advantages.
HBase scene
HBase is a database that can be said, it can be said to be a memory. HBase has a dual nature inherently have broad application scenarios. In some recent versions, introduced OffHeap gc reduce the impact and optimize link delay, provided Replica and so on to meet the online needs. The introduction of MOB, can store about 10M objects, fully adapted to the object store. In addition, as its concurrent power, storage capacity, it can be said to have the most competitive engine
-
Object Storage: We know a lot of headlines like, news of the news, web pages, images stored in HBase, some of the virus's virus database is stored in HBase
-
Time series data: on top of HBase has OpenTSDB modules, to meet the timing requirements of the class scene
-
Recommended portrait: in particular, the user's portrait, is a relatively large sparse matrix, risk control ants is built on top of HBase
-
Spatio-temporal data: main track, the weather grid and the like, drops a taxi trajectory data mainly in HBase, another large amount of data in all the little car networking technology companies, the data are in the presence of HBase
-
CubeDB OLAP: Kylin a cube analysis tools, the underlying data is stored in HBase, a lot of off-line calculation based on the customer's own building cube stored in the hbase, to meet the needs of the online report query
-
Message / order: in the field of telecommunications, the banking sector, many of the underlying storage order inquiry, in addition to a lot of communication, messaging application synchronization built on top of HBase
-
Feeds flow: typical application is the circle of friends xx similar applications
-
NewSQL: Phoenix plug on there to meet secondary indexes, SQL needs, the need butt traditional SQL data non-transactional needs
More scenes need to keep digging
The above review again HBase scene, made a simple classification, there will be some follow-up for some of the scenes of actual cases
Finally, welcome attention HBase Technical Community