[
https://issues.apache.org/jira/browse/GIRAPH-667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13659850#comment-13659850
]
Avery Ching commented on GIRAPH-667:
+1, but let's wait one or two days for others to
Hi All,
This year for the Nutch + Giraph LinkRank delegation and integration
proposal, we have two proposals currently sitting.
As a requirement, each proposal needs a backup mentor.
Please contact me off list if you are able to sign your name on the line as
a backup mentor in the event that I drop
Sorry, I guess I am bad at explaining things sometimes :-) MasterCompute can
decide to switch computation and combiner (I added methods setComputation and
setCombiner there) - so it's dynamic. Please take a look at TestSwitchClasses,
it's a dummy example but shows how these new stuff can be used
Ok, now I get it. I missed the part where the job would be a pipeline of
different computations, ala mapreduce. How do you register the next
computation in the pipeline? who's responsible for it? is it static prior
to the computation or dynamic and managed by some kind of master?
On Thu, May 16,
I don't think I'm following – how could we have three? And what does 'ad hoc
message value' mean? During a single superstep, we need to know what the
message type which we are sending is in order to be able to deserialize message
requests. And in the following superstep that same type has to be
It makes sense, but I guess an ad hoc message value would solve this. I'm a
bit shuffled by the idea of the number two. why not three? looks like an ad
hoc solution to me.
On Wed, May 15, 2013 at 9:13 PM, Maja Kabiljo wrote:
> Sure. In order to be able to send different message types during
>