hama-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Felix Halim <felix.ha...@gmail.com>
Subject Re: More documentation on Hama- BSP
Date Wed, 03 Feb 2010 08:54:57 GMT
On Wed, Feb 3, 2010 at 4:41 PM, Edward J. Yoon <edwardyoon@apache.org> wrote:
>>> I think the needToVisit() function might as well need to communicate
>>> with other machine:
> Hmm, You're exactly right. In that example, needToVisit() function
> checks the IsVisited from some shared-space (e.g., HBase or DBMS, ...,
> etc). We wrote with intent to simplify it.

>From the pseudocode, I see that every Vertex will request IsVisited.
In a large graph, the HBase or DBMS will be overwhelmed by many tiny
requests from each Vertex.
Does the needToVisit() has "bulk query" that aggregates the tiny
requests into a single request?

> Integer is the distance at "Map<Vertex, Integer> input, Map<Vertex,
> Integer> nextQueue", but it could be replaced as other object. for
> example, new PathWeHaveFollowed(). Then, perhaps we need not some
> shared-space.

But we also need to know PathOthersHaveFollowed() don't we?
That's why we need HBase to store the global "visited" states of each node.

Felix Halim

View raw message