[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-24 Thread Ilias Kapouranis (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13856383#comment-13856383
 ] 

Ilias Kapouranis commented on HAMA-833:
---

Yeah having this flexibility in aggregating will be great.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Edward J. Yoon (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13856063#comment-13856063
 ] 

Edward J. Yoon commented on HAMA-833:
-

Additionally, Anastatis, could you please explain about skipAggregator() method 
again? 

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Edward J. Yoon (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13856056#comment-13856056
 ] 

Edward J. Yoon commented on HAMA-833:
-

{quote}Exactly, but i mention again that this only happens at the very next 
aggregation phase after the vertex has halted. After that one, they are 
excluded as expected.{quote}

That's why I ask again. The "Aggregator" aggregates the value of *all* 
vertices[1]. It doesn't matter at all whether active or inactive. 
AverageAggregator should returns the average of all vertices, SumAggregator 
should returns the sum of all vertices. Min or Max is the same. Otherwise, it 
will affect the user's graph algorithm itself. The system bug I mentioned, is 
that aggregated value has changed.

1. 
http://svn.apache.org/repos/asf/hama/trunk/graph/src/main/java/org/apache/hama/graph/Aggregator.java

{quote}So, what's the status of this issue?{quote}

I don't see any relation between the reported issue and this ticket. I'd like 
to close this ticket.

Instead, to support aggregation for activated vertices, we can introduce the 
custom aggregator that can be used within Vertex like below: 

{code}

aggregate(MY_AGG, new DoubleWritable(1));
 } else {
voteToHalt();
 }
{code}

WDYT?

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Anastasis Andronidis (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855807#comment-13855807
 ] 

Anastasis Andronidis commented on HAMA-833:
---

So, what's the status of this issue? Re-open? And what about HAMA-807?

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Ilias Kapouranis (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855656#comment-13855656
 ] 

Ilias Kapouranis commented on HAMA-833:
---

Exactly, but i mention again that this only happens at the very next 
aggregation phase after the vertex has halted. After that one, they are 
excluded as expected.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Edward J. Yoon (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855653#comment-13855653
 ] 

Edward J. Yoon commented on HAMA-833:
-

OKay, I finally understood. The values of inactive vertex should be excluded 
from aggregation. Right?

And, system has a bug.
{code}
  if (this.getSuperstepCount() == 0) {
this.setValue(new DoubleWritable(1));
if(this.getVertexID().equals(new Text("2"))) {
  this.setValue(new DoubleWritable(0.5));
}
  } else if (this.getSuperstepCount() == 1) {
// do nothing
  } else if (this.getSuperstepCount() == 2) {
if(this.getVertexID().equals(new Text("2"))) {
  this.voteToHalt();
}
  } else if (this.getSuperstepCount() == 3) {
System.out.println("vertexID "+ this.getVertexID() +" : " + 
this.getLastAggregatedValue(0));
  }

Result:
vertexID 0 : 0.95
vertexID 4 : 0.95
vertexID 6 : 0.95
vertexID 8 : 0.95
vertexID 1 : 0.95
vertexID 3 : 0.95
vertexID 5 : 0.95
vertexID 7 : 0.95
vertexID 9 : 0.95
{code}

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Ilias Kapouranis (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855636#comment-13855636
 ] 

Ilias Kapouranis commented on HAMA-833:
---

Insted of getLastAggregatedValue(0) use getNumLastAggregatedVertices(0).

I know that he doesn't get called and this is right. But he gets aggregated at 
1 superstep!
My test code is the one pasted. And one line in which initializes the vertex 
values.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Edward J. Yoon (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855627#comment-13855627
 ] 

Edward J. Yoon commented on HAMA-833:
-

I've tested like you: 

{code}
  if (this.getSuperstepCount() == 0) {
this.setValue(new DoubleWritable(1));
  } else if (this.getSuperstepCount() == 1) {
if(this.getVertexID().equals(new Text("2"))) {
  this.voteToHalt();
}
  } else if (this.getSuperstepCount() == 2) {
System.out.println("vertexID "+ this.getVertexID() +" : " + 
this.getLastAggregatedValue(0));
  } else if (this.getSuperstepCount() == 3) {
System.out.println("vertexID "+ this.getVertexID() +" : " + 
this.getLastAggregatedValue(0));
  }

Result: 

vertexID 0 : 1.0
vertexID 4 : 1.0
vertexID 6 : 1.0
vertexID 8 : 1.0
vertexID 1 : 1.0
vertexID 3 : 1.0
vertexID 5 : 1.0
vertexID 7 : 1.0
vertexID 9 : 1.0

vertexID 0 : 1.0
vertexID 4 : 1.0
vertexID 6 : 1.0
vertexID 8 : 1.0
vertexID 1 : 1.0
vertexID 3 : 1.0
vertexID 5 : 1.0
vertexID 7 : 1.0
vertexID 9 : 1.0
{code}

Halted vertex doesn't called, and I don't see any problem. 

Can you attach your full test code? I don't know what are you talking about.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Ilias Kapouranis (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855608#comment-13855608
 ] 

Ilias Kapouranis commented on HAMA-833:
---

if(getSuperstepCount() == 1){
if(getVertexID().toString().equals("1")){
voteToHalt();
}
} else if(getSuperstepCount() == 2){
System.out.println(getNumLastAggregatedVertices(0));
} else if(getSuperstepCount() == 3){
System.out.println(getNumLastAggregatedVertices(0));
} 

So here we choose the vertex with id 1 to halt. At superstep 2 and 3 we print 
the number of vertices that were aggregated before each superstep.
If we insert 10 nodes, at superstep 2 we will see that it prints 10. This means 
that even though the node halted, it was aggregated! But then at superstep 3, 
we will see that the number goes to 9.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Edward J. Yoon (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855602#comment-13855602
 ] 

Edward J. Yoon commented on HAMA-833:
-

I don't know what are you talking about. Could you describe your simulation 
code here? and explain what's the problem?

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Ilias Kapouranis (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855598#comment-13855598
 ] 

Ilias Kapouranis commented on HAMA-833:
---

Imagine we are at superstep 3 and node K votes to halt. He WILL BE aggregated 
at the next aggregation phase. But from superstep 4 and onward he will be 
inactive.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Edward J. Yoon (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855581#comment-13855581
 ] 

Edward J. Yoon commented on HAMA-833:
-

As I mentioned above, inactive vertex is doing nothing until activated.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Ilias Kapouranis (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855567#comment-13855567
 ] 

Ilias Kapouranis commented on HAMA-833:
---

Yeah I mean the vertex that votes to halt, skips the very next aggregation 
phase.
In the code you posted, if a node votes to halt at a superstep gets aggregated 
without checking again!

A simple change would be:

 if (iterable != null && vertex.isHalted()) {
vertex.setActive();
 }
 if (!vertex.isHalted()) {
M lastValue = vertex.getValue();
if (iterable == null) {
   vertex.compute(Collections. emptyList());
} else {
   if (combiner != null) {
  M combined = combiner.combine(iterable);
  vertex.compute(Collections.singleton(combined));
   } else {
 vertex.compute(iterable);
  } 
  currentMessage = iterable.getOverflowMessage();
}   
}

if (!vertex.isHalted()) {
   getAggregationRunner().aggregateVertex(lastValue, vertex);
   activeVertices++;
}


With this we check again if the node has voted to halt in the same superstep.
I don't if this code will work but I think you get my idea.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Edward J. Yoon (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=1386#comment-1386
 ] 

Edward J. Yoon commented on HAMA-833:
-

With this, I'm very doubt about skipAggregator() method HAMA-807. How this 
works?

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Edward J. Yoon (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=1383#comment-1383
 ] 

Edward J. Yoon commented on HAMA-833:
-

Basically, inactive vertex is doing nothing until activated.  And aggregator is 
just a global message among vertices. 

{code}
  if (iterable != null && vertex.isHalted()) {
vertex.setActive();
  }
  if (!vertex.isHalted()) {
M lastValue = vertex.getValue();
if (iterable == null) {
  vertex.compute(Collections. emptyList());
} else {
  if (combiner != null) {
M combined = combiner.combine(iterable);
vertex.compute(Collections.singleton(combined));
  } else {
vertex.compute(iterable);
  }
  currentMessage = iterable.getOverflowMessage();
}
getAggregationRunner().aggregateVertex(lastValue, vertex);
activeVertices++;
  }
{code}



> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Ilias Kapouranis (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855544#comment-13855544
 ] 

Ilias Kapouranis commented on HAMA-833:
---

Well the only thing is that if a node votes to halt, the aggregation phase 
should be skipped.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Edward J. Yoon (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855541#comment-13855541
 ] 

Edward J. Yoon commented on HAMA-833:
-

I've commented on that thread. Maybe I'm not understand what you propose 
exactly but I think we have to fix a bug instead of adding these new features.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-23 Thread Ilias Kapouranis (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855522#comment-13855522
 ] 

Ilias Kapouranis commented on HAMA-833:
---

If you want to find the min or max (for example) value from the vertexes and 
continue with the ones that satisfy the condition, having the nodes being 
aggregated at the superstep they vote to halt hides some errors. I came up with 
the problem and we discussed it a bit at the user forum.
link : 
http://mail-archives.apache.org/mod_mbox/hama-user/201312.mbox/%[email protected]%3E

Specifically:
Your algorithm has 2 supersteps for each of its' iteration.
- At the first superstep assigns are done based on the aggregated values and 
messages are send.
- At the second nodes start voting to halt based on the messages received.

Now the problem here is that the nodes which voted to halt will be aggregated 
and will mess up the result for the next iterations.

Having the option to exclude them from the aggregation phase in which they 
voted to halt gives better control.

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)


[jira] [Commented] (HAMA-833) Add more finish states for vertices

2013-12-22 Thread Edward J. Yoon (JIRA)

[ 
https://issues.apache.org/jira/browse/HAMA-833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13855378#comment-13855378
 ] 

Edward J. Yoon commented on HAMA-833:
-

What's the use case scenario here?

> Add more finish states for vertices
> ---
>
> Key: HAMA-833
> URL: https://issues.apache.org/jira/browse/HAMA-833
> Project: Hama
>  Issue Type: Improvement
>  Components: graph
>Affects Versions: 0.6.3
>Reporter: Anastasis Andronidis
>Assignee: Anastasis Andronidis
>Priority: Minor
>  Labels: features
> Fix For: 0.7.0
>
>
> We should handle more cases on the vertices, like:
> 1) voteToStop() : Immediately stop the vertex compute and suppress any 
> further calculations on top of that. (e.g. aggregation)
> 2) voteToTerminate(): Immediately stop the vertex compute, suppress any 
> further calculations on top of that and deactivate the vertex so even if any 
> message reaches it, will not come alive.
> Any comments?



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)