[
https://issues.apache.org/jira/browse/HBASE-13485?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Andrew Purtell updated HBASE-13485:
-----------------------------------
Description:
Ceph (http://ceph.com/ceph-storage/) is an open source, scalable, distributed,
clustered block storage system with interesting dynamic data placement
properties and support for tiered storage and multiple strategies for
resilience and insuring data availability and integrity. Ceph is a layered
system: a block storage layer, known as RADOS, and a filesystem layer
implemented on top of the block store. The layers of Ceph have varying degrees
of stability. The block storage layer is claimed to be stable.
It might be interesting to try out HBase on Ceph using the Hadoop filesystem
binding for Ceph (https://github.com/ceph/cephfs-hadoop). Alternatively,
implement "HBase filesystem" support for RADOS via the Java bindings
(https://github.com/ceph/rados-java, via JNA). Considerations here will be
similar to what we might want to do when HDFS decouples block management from
namespace.
was:
Ceph (http://ceph.com/ceph-storage/ is an open source, scalable, distributed,
clustered block storage system with interesting dynamic data placement
properties and support for tiered storage and multiple strategies for
resilience and insuring data availability and integrity. Ceph is a layered
system: a block storage layer, known as RADOS, and a filesystem layer
implemented on top of the block store. The layers of Ceph have varying degrees
of stability. The block storage layer is claimed to be stable.
It might be interesting to try out HBase on Ceph using the Hadoop filesystem
binding for Ceph (https://github.com/ceph/cephfs-hadoop). Alternatively,
implement "HBase filesystem" support for RADOS via the Java bindings
(https://github.com/ceph/rados-java, via JNA). Considerations here will be
similar to what we might want to do when HDFS decouples block management from
namespace.
> HBase on Ceph
> -------------
>
> Key: HBASE-13485
> URL: https://issues.apache.org/jira/browse/HBASE-13485
> Project: HBase
> Issue Type: Task
> Reporter: Andrew Purtell
>
> Ceph (http://ceph.com/ceph-storage/) is an open source, scalable,
> distributed, clustered block storage system with interesting dynamic data
> placement properties and support for tiered storage and multiple strategies
> for resilience and insuring data availability and integrity. Ceph is a
> layered system: a block storage layer, known as RADOS, and a filesystem layer
> implemented on top of the block store. The layers of Ceph have varying
> degrees of stability. The block storage layer is claimed to be stable.
> It might be interesting to try out HBase on Ceph using the Hadoop filesystem
> binding for Ceph (https://github.com/ceph/cephfs-hadoop). Alternatively,
> implement "HBase filesystem" support for RADOS via the Java bindings
> (https://github.com/ceph/rados-java, via JNA). Considerations here will be
> similar to what we might want to do when HDFS decouples block management from
> namespace.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
