1
0
Fork 0
arangodb/arangod/Pregel
Simon d526805e81 Bug fix/fix suspicous stuff (#10273) 2019-10-17 15:34:22 +02:00
..
Algos
examples
Aggregator.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
AggregatorHandler.cpp
AggregatorHandler.h
AlgoRegistry.cpp
AlgoRegistry.h
Algorithm.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
CommonFormats.h
Conductor.cpp Bug fix/fix suspicous stuff (#10273) 2019-10-17 15:34:22 +02:00
Conductor.h
Graph.h
GraphFormat.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
GraphSerializer.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
GraphStore.cpp
GraphStore.h
IncomingCache.cpp C++17 (#10178) 2019-10-16 20:28:27 +03:00
IncomingCache.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
IndexHelpers.cpp
IndexHelpers.h
Iterators.h
MasterContext.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
MessageCombiner.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
MessageFormat.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
OutgoingCache.cpp Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
OutgoingCache.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
PregelFeature.cpp Feature/fuerte cluster 9000 (#10137) 2019-10-08 21:06:38 +02:00
PregelFeature.h
README.md
Recovery.cpp
Recovery.h
Statistics.h
TypedBuffer.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
Utils.cpp
Utils.h
VertexComputation.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
Worker-templates-algorithms.cpp
Worker-templates-native-types.cpp
Worker.cpp
Worker.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00
WorkerConfig.cpp Bug fix/fix suspicous stuff (#10273) 2019-10-17 15:34:22 +02:00
WorkerConfig.h
WorkerContext.h Use explicit default destructors where possible (#10166) 2019-10-04 15:58:30 +02:00

README.md

Pregel Subsystem

The Pregel subsystem implements a variety of different graph algorithms, this readme is more intended for internal use.

Protocol

Message format between DBServers:

{
  "sender": "someid",
  "executionNumber": 1337,
  "globalSuperstep": 123,
  "messages": [<vertexID1>, <slice1>, <vertexID2>, <slice2>]
}

Any type of slice is supported

Useful Commands

Import graph e.g. https://github.com/arangodb/example-datasets/tree/master/Graphs/1000 First rename the columns '_key', '_from', '_to' arangoimport will keep those.

In arangosh:

db._create('vertices', {numberOfShards: 2});
db._createEdgeCollection('alt_edges');
db._createEdgeCollection('edges', {numberOfShards: 2, shardKeys:["_vertex"], distributeShardsLike:'vertices'});
arangoimport --file generated_vertices.csv --type csv --collection vertices --overwrite true --server.endpoint http+tcp://127.0.0.1:8530

Or:

for(var i=0; i < 5000; i++) db.vertices.save({_key:i+""});
arangoimport --file generated_edges.csv --type csv --collection alt_edges --overwrite true --from-collection-prefix "vertices" --to-collection-prefix "vertices" --convert false --server.endpoint http+tcp://127.0.0.1:8530

AQL script to copy edge collection into one with '_vertex':

FOR doc IN alt_edges
  INSERT {
    _vertex: SUBSTRING(doc._from, FIND_FIRST(doc._from,"/") + 1),
    _from: doc._from,
    _to: doc._to
  } INTO edges
  LET values = (
    FOR s IN vertices
      RETURN s.result
  )
  RETURN SUM(values)

AWK Scripts

Make CSV file with IDs unique:

cat edges.csv | tr '[:space:]' '[\n*]' | grep -v "^\s*$" | awk '!seen[$0]++' > vertices.csv

Make CSV file with ArangoDB compatible edges:

cat edges.csv | awk -F" " '{print "profiles/" $1 "\tprofiles/" $2 "\t" $1}' >> arango-edges.csv

Import the vertex and edge CSV files:

arangoimport --file vertices.csv --type csv --collection twitter_v --overwrite true --convert false --server.endpoint http+tcp://127.0.0.1:8530  -c none
arangoimport --file arango-edges.csv --type csv --collection twitter_e --overwrite true --convert false --separator "\t" --server.endpoint http+tcp://127.0.0.1:8530  -c none