[
https://issues.apache.org/jira/browse/FLINK-954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14080917#comment-14080917
]
Jonathan Hasenburg edited comment on FLINK-954 at 7/31/14 2:06 PM:
-------------------------------------------------------------------
Where can I find information about the convergence criterion, here is an
example json created by flink from the current kmeans example... .
{code}
"nodes": [
{
"id": 2,
"type": "source",
"pact": "Data Source",
"contents": "[-14.22 -48.01, -22.78 37.1, 56.18 -42.99, 35.04
50.29, -9.53 -46.26, -34.35 48.25, 55.82 -57.49, 21",
"parallelism": "1",
"subtasks_per_instance": "1"
},
{
"id": 4,
"type": "source",
"pact": "Data Source",
"contents": "[1 -31.85 -44.77, 2 35.16 17.46, 3 -5.16 21.93, 4
-24.06 6.81]",
"parallelism": "1",
"subtasks_per_instance": "1"
},
{
"step_function": [
{
"id": 9,
"type": "pact",
"pact": "Bulk Partial Solution",
"contents": "Partial Solution",
"parallelism": "1",
"subtasks_per_instance": "1"
},
{
"id": 8,
"type": "pact",
"pact": "Map",
"contents":
"eu.stratosphere.example.java.clustering.KMeans$SelectNearestCenter",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 2, "side": "first"},
{"id": 9, "side": "second", "ship_strategy":
"Broadcast"}
]
},
{
"id": 7,
"type": "pact",
"pact": "Map",
"contents":
"eu.stratosphere.example.java.clustering.KMeans$CountAppender",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 8}
]
},
{
"id": 6,
"type": "pact",
"pact": "Reduce",
"contents":
"eu.stratosphere.example.java.clustering.KMeans$CentroidAccumulator",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 7}
]
},
{
"id": 5,
"type": "pact",
"pact": "Map",
"contents":
"eu.stratosphere.example.java.clustering.KMeans$CentroidAverager",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 6}
]
}
],
"partial_solution": 9,
"next_partial_solution": 5,
"id": 3,
"type": "bulk_iteration",
"pact": "Bulk Iteration",
"contents": "Bulk Iteration",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 4}
]
},
{
"id": 1,
"type": "pact",
"pact": "Map",
"contents":
"eu.stratosphere.example.java.clustering.KMeans$SelectNearestCenter",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 2, "side": "first"},
{"id": 3, "side": "second", "ship_strategy":
"Broadcast"}
]
},
{
"id": 0,
"type": "sink",
"pact": "Data Sink",
"contents": "Print to System.out",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 1}
]
}
]
}
}
{code}
was (Author: jonathanh5):
Where can I find information about the convergence criterion, here is an
example json create by flink from the current kmeans example... .
{code}
"nodes": [
{
"id": 2,
"type": "source",
"pact": "Data Source",
"contents": "[-14.22 -48.01, -22.78 37.1, 56.18 -42.99, 35.04
50.29, -9.53 -46.26, -34.35 48.25, 55.82 -57.49, 21",
"parallelism": "1",
"subtasks_per_instance": "1"
},
{
"id": 4,
"type": "source",
"pact": "Data Source",
"contents": "[1 -31.85 -44.77, 2 35.16 17.46, 3 -5.16 21.93, 4
-24.06 6.81]",
"parallelism": "1",
"subtasks_per_instance": "1"
},
{
"step_function": [
{
"id": 9,
"type": "pact",
"pact": "Bulk Partial Solution",
"contents": "Partial Solution",
"parallelism": "1",
"subtasks_per_instance": "1"
},
{
"id": 8,
"type": "pact",
"pact": "Map",
"contents":
"eu.stratosphere.example.java.clustering.KMeans$SelectNearestCenter",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 2, "side": "first"},
{"id": 9, "side": "second", "ship_strategy":
"Broadcast"}
]
},
{
"id": 7,
"type": "pact",
"pact": "Map",
"contents":
"eu.stratosphere.example.java.clustering.KMeans$CountAppender",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 8}
]
},
{
"id": 6,
"type": "pact",
"pact": "Reduce",
"contents":
"eu.stratosphere.example.java.clustering.KMeans$CentroidAccumulator",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 7}
]
},
{
"id": 5,
"type": "pact",
"pact": "Map",
"contents":
"eu.stratosphere.example.java.clustering.KMeans$CentroidAverager",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 6}
]
}
],
"partial_solution": 9,
"next_partial_solution": 5,
"id": 3,
"type": "bulk_iteration",
"pact": "Bulk Iteration",
"contents": "Bulk Iteration",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 4}
]
},
{
"id": 1,
"type": "pact",
"pact": "Map",
"contents":
"eu.stratosphere.example.java.clustering.KMeans$SelectNearestCenter",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 2, "side": "first"},
{"id": 3, "side": "second", "ship_strategy":
"Broadcast"}
]
},
{
"id": 0,
"type": "sink",
"pact": "Data Sink",
"contents": "Print to System.out",
"parallelism": "1",
"subtasks_per_instance": "1",
"predecessors": [
{"id": 1}
]
}
]
}
}
{code}
> Improvements to the Web Frontend Plan presentation
> --------------------------------------------------
>
> Key: FLINK-954
> URL: https://issues.apache.org/jira/browse/FLINK-954
> Project: Flink
> Issue Type: Improvement
> Components: Webfrontend
> Affects Versions: 0.6-incubating
> Reporter: Stephan Ewen
> Assignee: Jonathan Hasenburg
> Priority: Minor
>
> I propose some changes to the way the plan is displayed in the webfrontend:
> 1. Add zoom buttons (+/-). On Laptops where you have no mouse wheel, you
> cannot zoom the graph.
> 2. Omit the ID in the title of the operators (maybe list it in the bottom)
> 3. Rename "Driver Strategy" to "Strategy" or "Operations"
> 4. Make the arrows from Broadcast variables enter the box from top/bottom,
> rather than from the left.
> 5. Instead of writing in a node "Additional Edge to Node X" , can you
> actually draw a box X and have an arrow from X to that node?
--
This message was sent by Atlassian JIRA
(v6.2#6252)
