Http://www.aliyun.com/zixun/aggregation/1560.html ">facebook recently on the official blog announced the introduction of the HBase Database upgrade version of--hydrabase, Facebook is a heavy user of hbase, and Facebook's HBase database system stores Many of Facebook's key business data, including internal monitoring systems, search indexing, streaming data analysis, and data capture. Hydrabase can reduce server downtime when compared to hbase stability and availability.
Facebook says Hydrabase can reduce Facebook's annual downtime to less than 5 minutes.
In HBase systems, data fragmentation is stored in many areas, and if a zone server goes down, its domain data needs to be migrated to another domain server. Facebook points out that although HBase can recover automatically, the recovery time is too long.
Typical deployment model for Hydrabase
Hydrabase enables a data domain to be distributed across multiple domain servers, with mutual backup between domain servers, which can greatly reduce the time spent in data recovery. Facebook says Hydrabase can reduce Facebook's annual downtime to less than 5 minutes.
Facebook is currently testing hydrabase and is planning to start deploying gradually in the production cluster. Outside of Hydrabase, Facebook also unveiled details of HDFs RAID and its error-correcting code for data protection to streamline the number of clusters that create backups with Hadoop.
Last year Facebook used HDFs RAID technology in the Data Warehouse cluster for the first time, with a significant increase in backup efficiency and a savings of 10 petabytes of data storage space.