Spark GraphX: add multiple edge weights

Question

I am new to GraphX and have a Spark dataframe with four columns like below:

src_ip    dst_ip    flow_count   sum_bytes
8.8.8.8   1.2.3.4          435        1137
  ...       ...           ...         ...

Basically I want to map both src_ip and dst_ip to vertices and assign flow_count and sum_bytes as edges attribute. As far as I know, we can not add edges attributes in GraphX as only vertex attributes are permitted. Hence, I am thinking about adding flow_count as edge weight:

//create edges
val trafficEdges = trafficsFromTo.map(x =Edge(MurmurHash3.stringHash(x(0).toString,MurmurHash3.stringHash(x(1).toString,x(2))

However, can I add sum_bytes as edge weight as well?

Answer 1

It is possible to add both variables to the edge. The simplest solution would be to use a tuple, for example:

val data = Array(Edge(3L, 7L, (123, 456)), Edge(5L, 3L, (41, 34)))
val edges: RDD[Edge[(Int, Int)]] = spark.sparkContext.parallelize(data)

Alternatively, you can make use of a case class:

case class EdgeWeight(flow_count: Int, sum_bytes: Int)

val data2 = Array(Edge(3L, 7L, EdgeWeight(123, 456)), Edge(5L, 3L, EdgeWeight(41, 34)))
val edges: RDD[Edge[EdgeWeight]] = spark.sparkContext.parallelize(data2)

Using a case class would be more convenient to use and maintain if there are more attributes to be added.

---

I believe that in this specific case, it is most elegantly solved by:

val trafficEdges = trafficsFromTo.map{x => 
  Edge(MurmurHash3.stringHash(x(0).toString, 
       MurmurHash3.stringHash(x(1).toString,
       EdgeWeight(x(2), x(3))
}

trafficEdges.sortBy(edge => edge.attr.flow_count) // sort by flow_count