http://www.datastax.com/dev/blog/cassandra-file-system-design - The
gist is that Datastax has replaced the Namenode, Secondary Namenode,
and (I forget the other) in order to eliminate the single point of
failure. This raises the strange notion of running Accumulo on top of
Cassandra.

I believe Eric has been working on something very similar using Accumulo
itself: https://issues.apache.org/jira/browse/ACCUMULO-722

On 10/1/2012 9:42 PM, David Medinets wrote:
> http://www.datastax.com/dev/blog/cassandra-file-system-design - The
> gist is that Datastax has replaced the Namenode, Secondary Namenode,
> and (I forget the other) in order to eliminate the single point of
> failure. This raises the strange notion of running Accumulo on top of
> Cassandra.
>

If SPOF is your concern, why not just use Hadoop 2?

-Todd

On Mon, Oct 1, 2012 at 6:42 PM, David Medinets <[EMAIL PROTECTED]> wrote:
> http://www.datastax.com/dev/blog/cassandra-file-system-design - The
> gist is that Datastax has replaced the Namenode, Secondary Namenode,
> and (I forget the other) in order to eliminate the single point of
> failure. This raises the strange notion of running Accumulo on top of
> Cassandra.

--
Todd Lipcon
Software Engineer, Cloudera

I think the hope is that the Cassandra File System can be used to
replicate between data centers. For example, two replicas in center A
and two replicas in center B.

On Mon, Oct 1, 2012 at 10:34 PM, Todd Lipcon <[EMAIL PROTECTED]> wrote:
> If SPOF is your concern, why not just use Hadoop 2?
>
> -Todd
>
> On Mon, Oct 1, 2012 at 6:42 PM, David Medinets <[EMAIL PROTECTED]> wrote:
>> http://www.datastax.com/dev/blog/cassandra-file-system-design - The
>> gist is that Datastax has replaced the Namenode, Secondary Namenode,
>> and (I forget the other) in order to eliminate the single point of
>> failure. This raises the strange notion of running Accumulo on top of
>> Cassandra.
>
>
>
> --
> Todd Lipcon
> Software Engineer, Cloudera