[ https://issues.apache.org/jira/browse/OMID-90?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16543619#comment-16543619 ]
Lianghong Xu commented on OMID-90: ---------------------------------- Hi [~ohads] [~ebortnik] I just did an initial evaluation with Omid Low Latency version with YCSB. However, the result in terms of throughput and latency seems similar to the original Omid. I guess maybe there is something I'm missing? I was using the default configs except that I set "lowLatency: true" to be true in omid-server-configuration.yml. I think that should be enough to turn on the low latency mode, right? The YCSB workload we used has a read/write ratio of 80/20 with uniform dist. Please let us know if you have any suggestions on tuning. Thanks, Lianghong > Reducing begin/commit latency by distributing the write to the commit table > --------------------------------------------------------------------------- > > Key: OMID-90 > URL: https://issues.apache.org/jira/browse/OMID-90 > Project: Apache Omid > Issue Type: New Feature > Reporter: Ohad Shacham > Assignee: Yonatan Gottesman > Priority: Major > Attachments: OmidCloud-VLDB.pdf > > > Today, Omid's commits are done by the transaction manager. In order to > efficiently write to the commit table, the transaction manager batches these > writes. This optimization, even thought reduces the write time to HBase, > significantly increases the begin and commit latency. The commit latency > increases since a commit operation returns only after its commit timestamp > was persisted in the commit table. And the begin latency increases since > begin returns a transaction id that is also used by the transaction to > identify its snapshot and therefore, begin returns only after all commits > with commit id smaller than the begin id was persisted in the commit table. > This is crucial, since a snapshot change during a transaction run may violate > snapshot isolation. > > The idea of this feature is to distribute the commit by moving the write to > the commit table from the server to the client. The transaction manager does > conflict analysis and returns a commit timestamp. While the client atomically > persists this commit in the commit table. > This significantly reduces the begin and commit latency, since batching is > not required anymore. A begin operation can immediately returns and a commit > operation returns after conflict detection. > This can introduce snapshot isolation violation since a slow client can > commit and change other transaction's snapsho. Therefore, we use an > invalidation technique which is similar to the one Omid uses today to > maintain snapshot isolation in high availability mode. -- This message was sent by Atlassian JIRA (v7.6.3#76005)