[
https://issues.apache.org/jira/browse/HBASE-2315?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12844759#action_12844759
]
Benjamin Reed commented on HBASE-2315:
--------------------------------------
to get performance you need a dedicated drive, not necessarily dedicated
hardware. bk takes care of the batching for you. there is opportunistic
batching that happens in the network channel and to the disk. (the later being
the most important.)
i don't think ryan's persistence question has been answered. when bk responds
successfully to a write request, it means that the data has been synced to the
requisite number of devices. when testing we usually use 3 or 5 bookies with a
replication factor of 2 or 3. (bookkeeper does striping.)
we have an asynchronous write interface to allow high throughput by pipelining
requests, we invoke completions as each write completes successfully.
> BookKeeper for write-ahead logging
> ----------------------------------
>
> Key: HBASE-2315
> URL: https://issues.apache.org/jira/browse/HBASE-2315
> Project: Hadoop HBase
> Issue Type: New Feature
> Components: regionserver
> Reporter: Flavio Paiva Junqueira
> Attachments: HBASE-2315.patch, zookeeper-dev-bookkeeper.jar
>
>
> BookKeeper, a contrib of the ZooKeeper project, is a fault tolerant and high
> throughput write-ahead logging service. This issue provides an implementation
> of write-ahead logging for hbase using BookKeeper. Apart from expected
> throughput improvements, BookKeeper also has stronger durability guarantees
> compared to the implementation currently used by hbase.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
