[
https://issues.apache.org/jira/browse/HDFS-11361?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chen Liang updated HDFS-11361:
------------------------------
Description:
This JIRA includes two major pieces: 1. the interface exposed to upper layer
operating system that mounts the file system; and 2. the interface exposed to
lower layer container storage layer.
On one hand, to make CBlock mountable and accessible for a POSIX OS, we need to
implement file system protocol. We chose to do this by running cblock with
jSCSI (http://jscsi.org/) which is a java implementation of iSCSI protocol.
More specifically, CBlock volume will exposed as an java-based iSCSI target,
and any machine that has iSCSI installed will be able to talk to CBlock volume
and mounted locally just as any disks. This is JIRA make a starting point by
introducing jSCSI and implementing its {{IStorageModule}}. This will be
registered to jSCSI server in future JIRA.
One the other hand, to be able to have the CBlock talk to container layer with
better performance, CBlock should be able to cache data locally. More
specifically, clients' read/write should be performed on the locally disk
cache, while threads are pushing and retrieving data in backend. This JIRA
takes the first step by defining interfaces for the local cache.
was:
This JIRA adds two major pieces: 1. the interface exposed to upper layer
operating system that mounts the file system; and 2. the interface exposed to
lower layer container storage layer.
On one hand, to make CBlock mounted and accessible for OS, we need to implement
some protocol. We chose to do this by running cblock as a iSCSI
(http://jscsi.org/) which is a java implementation of iSCSI protocol. More
specifically, CBlock volume will exposed as an iSCSI target. This is JIRA make
a starting point by introducing jSCSI and implementing its {{IStorageModule}}.
One the other hand, to be able to talk to container layer with better
performance, cblock should be able to cache data locally. More
specifically,clients' read/write should be performed on the locally disk cache,
while threads are pushing and retrieving data in backend. This JIRA takes the
first step by defining interfaces for the local cache. This allows us to
implement caches with a baseline and makes the unit tests a lot easier.
> Block Storage : add cache interface
> -----------------------------------
>
> Key: HDFS-11361
> URL: https://issues.apache.org/jira/browse/HDFS-11361
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: hdfs
> Reporter: Chen Liang
> Assignee: Chen Liang
>
> This JIRA includes two major pieces: 1. the interface exposed to upper layer
> operating system that mounts the file system; and 2. the interface exposed to
> lower layer container storage layer.
> On one hand, to make CBlock mountable and accessible for a POSIX OS, we need
> to implement file system protocol. We chose to do this by running cblock with
> jSCSI (http://jscsi.org/) which is a java implementation of iSCSI protocol.
> More specifically, CBlock volume will exposed as an java-based iSCSI target,
> and any machine that has iSCSI installed will be able to talk to CBlock
> volume and mounted locally just as any disks. This is JIRA make a starting
> point by introducing jSCSI and implementing its {{IStorageModule}}. This will
> be registered to jSCSI server in future JIRA.
> One the other hand, to be able to have the CBlock talk to container layer
> with better performance, CBlock should be able to cache data locally. More
> specifically, clients' read/write should be performed on the locally disk
> cache, while threads are pushing and retrieving data in backend. This JIRA
> takes the first step by defining interfaces for the local cache.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
