Hi guys I've tested the write performance on both master branch and HBASE-19064 branch. Here's the JIRA [1].
The conclusion(Pasted from the comment): The QPS & AVG latency are similar when comparing the master branch with async replication and HBASE-19064 with async replication. BTW the QPS of HBASE-19064 with sync replication dropped about 13%, we will continue to optimize the sync replication in phrase#2 [2]. 1. https://issues.apache.org/jira/browse/HBASE-20751 2. https://issues.apache.org/jira/browse/HBASE-20422 On Thu, Jun 21, 2018 at 9:57 PM, Guanghao Zhang <zghao...@gmail.com> wrote: > +1 > > 2018-06-21 18:26 GMT+08:00 mengli721014 on 163 dot com < > mengli721...@163.com > >: > > > 0 > > > > 使用AquaMail for Android发送 > > http://www.aqua-mail.com > > > > > > > > 在 2018年6月21日 下午4:21:26 "张铎(Duo Zhang)" <palomino...@gmail.com> 写道: > > > > In HBASE-19064 we aim to implement sync replication feature for HBase. > You > >> can see the design doc for more details on how it works > >> > >> https://docs.google.com/document/d/193D3aOxD-muPIZuQfI4Zo3_ > >> qg6-Nepeu_kraYJVQkiE/edit# > >> > >> And now the feature basically works, we have tested it on real clusters, > >> and also provide the operational documentation in the ref guide. You can > >> see the 'Synchronous Replication' section in the ref guide of branch > >> HBASE-19064. > >> > >> There are known limitations for this feature, and we have > >> created HBASE-20422 to track them. > >> > >> Please vote: > >> [+1] Agree > >> [-1] Disagree > >> [0] Neutral > >> > >> Thanks. > >> > >> > > > > > > >