Now The Isolation and allocation part of HBase is available.
The issue is discussed in https://issues.apache.org/jira/browse/HBASE-4120,the 
source code and patched tarball are located in 
https://github.com/ICT-Ope/HBase_allocation, Two versions Integrated with HBase 
0.90.2 and 0.90.3 are available to use.
The HBase isolation and allocation tool is designed to help users manage 
cluster resource among different applications and tables.When we have a large 
scale of HBase cluster with many applications running on it, there will be lots 
of problems. In Taobao.Inc there is a cluster for many departments to test 
their applications' performance, these applications are based on HBase. With 
one cluster which 
has 12 servers, there will be only one application running exclusively on this 
server, many other applications must wait until the previous test 
finished.After we add allocation management function to the cluster, 
applications can share the cluster and run concurrently. Also if the Test 
Engineers want to make 
sure that there is no interference, they can move other tables out of this 
group.
Different groups can have different region server configurations, which means 
that some groups optimized for reading can have large block cache size, and 
others optimized for writing can have large memstore size. Tables and region 
servers can be moved easily between groups. After changing the configuration, a 
group can be restarted alone instead of restarting the 
whole cluster.
In groups we use table priority to allocate resource. When system is busy, this 
can make sure high-priority tables not affected by lower-priority ones
Though we put group and priority of tables together, users can use them 
separately, there is no inseparable link between these two functions.
Suggestions and advice are welcome, please send mails to me or post your 
opinions on JIRA.
Jia Liu
Key Laboratory of Network Science and Technology,
Institute of Computing Technology,
Chinese Academy of Sciences,
P.O.Box 2704, Beijing 100190, China.

Reply via email to