Scala

How to Get Started Using Apache Spark GraphX with Scala

Editor’s Note: Don’t miss our new free on-demand training course about how to create data pipeline applications using Apache Spark – learn more here.

This post will help you get started using Apache Spark GraphX with Scala on the MapR Sandbox. GraphX is the Apache Spark component for graph-parallel computations, built upon a branch of mathematics called graph theory. It is a distributed graph processing framework that sits on top of the Spark core.

Overview of some graph concepts

A graph is a mathematical structure used to model relations between objects. A graph is made up of vertices and edges that connect them. The vertices are the objects and the edges are the relationships between them.

image00_edge-vertex-relationship

A directed graph is a graph where the edges have a direction associated with them. An example of a directed graph is a Twitter follower. User Bob can follow user Carol without implying that user Carol follows user Bob.

image02_bob-follows-carol

A regular graph is a graph where each vertex has the same number of edges. An example of a regular graph is Facebook friends. If Bob is a friend of Carol, then Carol is also a friend of Bob.

GraphX Property Graph

GraphX extends the Spark RDD with a Resilient Distributed Property Graph.

The property graph is a directed multigraph which can have multiple edges in parallel. Every edge and vertex has user defined properties associated with it. The parallel edges allow multiple relationships between the same vertices.

image01_flight-relationship

In this activity, you will use GraphX to analyze flight data.

Scenario

As a starting simple example, we will analyze three flights. For each flight, we have the following information:

Originating AirportDestination AirportDistance
SFOORD1800 miles
ORDDFW>800 miles
DFWSFO>1400 miles

 

In this scenario, we are going to represent the airports as vertices and routes as edges. For our graph we will have three vertices, each representing an airport. The distance between the airports is a route property, as shown below:image04_3-vertex-relationship

Vertex Table for Airports

IDProperty
1SFO
2ORD
3DFW

 

Edges Table for Routes

SrcIdDestIdProperty
121800
23800
311400

 

Software

This tutorial will run on the MapR Sandbox, which includes Spark.

Launch the Spark Interactive Shell

Log into the MapR Sandbox, as explained in Getting Started with Spark on MapR Sandbox, using userid user01, password mapr. Start the spark shell with:

$ spark-shell

Define Vertices

First we will import the GraphX packages.

(In the code boxes, comments are in Green and output is in Blue)

import org.apache.spark._
import org.apache.spark.rdd.RDD
// import classes required for using GraphX
import org.apache.spark.graphx._

We define airports as vertices. Vertices have an Id and can have properties or attributes associated with them. Each vertex consists of :

  • Vertex id → Id (Long)
  • Vertex Property → name (String)

Vertex Table for Airports

IDProperty(V)
1SFO

 

We define an RDD with the above properties that is then used for the vertexes.

// create vertices RDD with ID and Name
val vertices=Array((1L, ("SFO")),(2L, ("ORD")),(3L,("DFW")))
val vRDD= sc.parallelize(vertices)
vRDD.take(1)
// Array((1,SFO)) 
// Defining a default vertex called nowhere
val nowhere = "nowhere"

Define Edges

Edges are the routes between airports. An edge must have a source, a destination, and can have properties. In our example, an edge consists of:

  • Edge origin id → src (Long)
  • Edge destination id → dest (Long)
  • Edge Property distance → distance (Long)

Edges Table for Routes

srciddestidProperty(E)
1121800

 

We define an RDD with the above properties that is then used for the edges. The edge RDD has the form (src id, dest id, distance ).

// create routes RDD with srcid, destid, distance
val edges = Array(Edge(1L,2L,1800),Edge(2L,3L,800),Edge(3L,1L,1400))
val eRDD= sc.parallelize(edges) 
eRDD.take(2)
// Array(Edge(1,2,1800), Edge(2,3,800))

Create Property Graph

To create a graph, you need to have a Vertex RDD, Edge RDD, and a Default vertex.

Create a property graph called graph.

// define the graph
val graph = Graph(vRDD,eRDD, nowhere)
// graph vertices
graph.vertices.collect.foreach(println)
// (2,ORD)
// (1,SFO)
// (3,DFW) 
// graph edges
graph.edges.collect.foreach(println) 
// Edge(1,2,1800)
// Edge(2,3,800)
// Edge(3,1,1400)

1. How many airports are there?

// How many airports?
val numairports = graph.numVertices
// Long = 3

2. How many routes are there?

// How many routes?
val numroutes = graph.numEdges
// Long = 3

3. which routes > 1000 miles distance?

// routes > 1000 miles distance?
graph.edges.filter { case Edge(src, dst, prop) => prop > 1000 }.collect.foreach(println)
// Edge(1,2,1800)
// Edge(3,1,1400)

4. The EdgeTriplet class extends the Edge class by adding the srcAttr and dstAttr members which contain the source and destination properties, respectively.

// triplets
graph.triplets.take(3).foreach(println)
((1,SFO),(2,ORD),1800)
((2,ORD),(3,DFW),800)
((3,DFW),(1,SFO),1400)

5. Sort and print out the longest distance routes

// print out longest routes
graph.triplets.sortBy(_.attr, ascending=false).map(triplet =>
     "Distance " + triplet.attr.toString + " from " + triplet.srcAttr + " to " + triplet.dstAttr + ".").collect.foreach(println) 
Distance 1800 from SFO to ORD.
Distance 1400 from DFW to SFO.
Distance 800 from ORD to DFW.

Analyze Real Flight Data with GraphX

Scenario

Our data is from http://www.transtats.bts.gov/DL_SelectFields.asp?Table_ID=236&DB_Short_Name=On-Time. We are using flight information for January 2015. For each flight, we have the following information:

FieldDescriptionExample Value
dOfM(String)Day of month1
dOfW (String)Day of week4
carrier (String)Carrier codeAA
tailNum (String)Unique identifier for the plane – tail numberN787AA
flnum(Int)Flight number21
org_id(String)Origin airport ID12478
origin(String)Origin Airport CodeJFK
dest_id (String)Destination airport ID12892
dest (String)Destination airport codeLAX
crsdeptime(Double)Scheduled departure time900
deptime (Double)Actual departure time855
depdelaymins (Double)Departure delay in minutes0
crsarrtime (Double)Scheduled arrival time1230
arrtime (Double)Actual arrival time1237
arrdelaymins (Double)Arrival delay minutes7
crselapsedtime (Double)Elapsed time390
dist (Int)Distance2475

 

In this scenario, we are going to represent the airports as vertices and routes as edges. We are interested in visualizing airports and routes and would like to see the number of airports that have departures or arrivals.

Log into the MapR Sandbox, as explained in Getting Started with Spark on MapR Sandbox, using userid user01, password mapr. Copy the sample data file rita2014jan.csv to your sandbox home directory /user/user01 using scp.

Start the Spark shell with:

$ spark-shell

Define Vertices

First we will import the GraphX packages.

(In the code boxes, comments are in Green and output is in Blue)

import org.apache.spark._
import org.apache.spark.rdd.RDD
import org.apache.spark.util.IntParam
// import classes required for using GraphX
import org.apache.spark.graphx._
import org.apache.spark.graphx.util.GraphGenerators

Below we use Scala case classes to define the flight schema corresponding to the csv data file.

// define the Flight Schema
case class Flight(dofM:String, dofW:String, carrier:String, tailnum:String, flnum:Int, org_id:Long, origin:String, dest_id:Long, dest:String, crsdeptime:Double, deptime:Double, depdelaymins:Double, crsarrtime:Double, arrtime:Double, arrdelay:Double,crselapsedtime:Double,dist:Int)

The function below parses a line from the data file into the flight class.

// function to parse input into Flight class
def parseFlight(str: String): Flight = {
 val line = str.split(",")
 Flight(line(0), line(1), line(2), line(3), line(4).toInt, line(5).toLong, line(6), line(7).toLong, line(8), line(9).toDouble, line(10).toDouble, line(11).toDouble, line(12).toDouble, line(13).toDouble, line(14).toDouble, line(15).toDouble, line(16).toInt)
}

Below we load the data from the csv file into a Resilient Distributed Dataset (RDD). RDDs can have transformations and actions, the first() action returns the first element in the RDD.

// load the data into a RDD
val textRDD = sc.textFile("/user/user01/data/rita2014jan.csv")
// MapPartitionsRDD[1] at textFile 
// parse the RDD of csv lines into an RDD of flight classes
val flightsRDD = textRDD.map(parseFlight).cache()

We define airports as vertices. Vertices can have properties or attributes associated with them. Each vertex has the following property:

  • Airport name (String)

Vertex Table for Airports

IDProperty(V)
10397ATL

 
We define an RDD with the above properties that is then used for the vertexes.

// create airports RDD with ID and Name
val airports = flightsRDD.map(flight => (flight.org_id, flight.origin)).distinct 
airports.take(1)
// Array((14057,PDX)) 
// Defining a default vertex called nowhere
val nowhere = "nowhere" 
// Map airport ID to the 3-letter code to use for printlns
val airportMap = airports.map { case ((org_id), name) => (org_id -> name) }.collect.toList.toMap
// Map(13024 -> LMT, 10785 -> BTV,…)

Define Edges

Edges are the routes between airports. An edge must have a source, a destination, and can have properties. In our example, an edge consists of:

  • Edge origin id → src (Long)
  • Edge destination id → dest (Long)
  • Edge property distance → distance (Long)

Edges Table for Routes

srciddestidProperty(E)
14869146831087

 

We define an RDD with the above properties that is then used for the edges. The edge RDD has the form (src id, dest id, distance).

// create routes RDD with srcid, destid, distance
val routes = flightsRDD.map(flight => ((flight.org_id, flight.dest_id), flight.dist)).distinctdistinct 
routes.take(2)
// Array(((14869,14683),1087), ((14683,14771),1482)) 
// create edges RDD with srcid, destid , distance
val edges = routes.map {
 case ((org_id, dest_id), distance) =>Edge(org_id.toLong, dest_id.toLong, distance) } 
edges.take(1)
//Array(Edge(10299,10926,160))

Create Property Graph

To create a graph, you need to have a Vertex RDD, Edge RDD and a Default vertex.

Create a property graph called graph.

// define the graph
val graph = Graph(airports, edges, nowhere) 
// graph vertices
graph.vertices.take(2)
Array((10208,AGS), (10268,ALO)) 
// graph edges
graph.edges.take(2)
Array(Edge(10135,10397,692), Edge(10135,13930,654))

6. How many airports are there?

// How many airports?
val numairports = graph.numVertices
// Long = 301

7. How many routes are there?

// How many airports?
val numroutes = graph.numEdges
// Long = 4090

8. Which routes > 1000 miles distance?

// routes > 1000 miles distance?
graph.edges.filter { case ( Edge(org_id, dest_id,distance))=> distance > 1000}.take(3)
// Array(Edge(10140,10397,1269), Edge(10140,10821,1670), Edge(10140,12264,1628))

9. The EdgeTriplet class extends the edge class by adding the srcAttr and dstAttr members which contain the source and destination properties, respectively.

// triplets
graph.triplets.take(3).foreach(println)
((10135,ABE),(10397,ATL),692)
((10135,ABE),(13930,ORD),654)
((10140,ABQ),(10397,ATL),1269)

10. Sort and print out the longest distance routes

// print out longest routes
graph.triplets.sortBy(_.attr, ascending=false).map(triplet =>
     "Distance " + triplet.attr.toString + " from " + triplet.srcAttr + " to " + triplet.dstAttr + ".").take(10).foreach(println) 
Distance 4983 from JFK to HNL.
Distance 4983 from HNL to JFK.
Distance 4963 from EWR to HNL.
Distance 4963 from HNL to EWR.
Distance 4817 from HNL to IAD.
Distance 4817 from IAD to HNL.
Distance 4502 from ATL to HNL.
Distance 4502 from HNL to ATL.
Distance 4243 from HNL to ORD.
Distance 4243 from ORD to HNL.

11. Compute the highest degree vertex

// Define a reduce operation to compute the highest degree vertex
def max(a: (VertexId, Int), b: (VertexId, Int)): (VertexId, Int) = {
 if (a._2 > b._2) a else b
}
val maxInDegree: (VertexId, Int) = graph.inDegrees.reduce(max)
//maxInDegree: (org.apache.spark.graphx.VertexId, Int) = (10397,152) 
val maxOutDegree: (VertexId, Int) = graph.outDegrees.reduce(max)
//maxOutDegree: (org.apache.spark.graphx.VertexId, Int) = (10397,153) 
val maxDegrees: (VertexId, Int) = graph.degrees.reduce(max)
//maxDegrees: (org.apache.spark.graphx.VertexId, Int) = (10397,305) 
// Get the name for the airport with id 10397
airportMap(10397)
//res70: String = ATL

12. Which airport has the most incoming flights?

// get top 3
val maxIncoming = graph.inDegrees.collect.sortWith(_._2 > _._2).map(x => (airportMap(x._1), x._2)).take(3) 
maxIncoming.foreach(println)
(ATL,152)
(ORD,145)
(DFW,143) 
// which airport has the most outgoing flights?
val maxout= graph.outDegrees.join(airports).sortBy(_._2._1, ascending=false).take(3) 
maxout.foreach(println)
(10397,(153,ATL))
(13930,(146,ORD))
(11298,(143,DFW))

PageRank

Another GraphX operator is PageRank. which is based on the Google PageRank algorithm.

PageRank measures the importance of each vertex in a graph, by determining which vertexes have the most edges with other vertexes. In our example, we can use PageRank to determine which airports are the most important by measuring which airports have the most connections to other airports.

We have to specify the tolerance, which is the measure of convergence.

13. What are the most important airports according to PageRank?

// use pageRank
val ranks = graph.pageRank(0.1).vertices
// join the ranks  with the map of airport id to name
val temp= ranks.join(airports)
temp.take(1)
// Array((15370,(0.5365013694244737,TUL))) 
// sort by ranking
val temp2 = temp.sortBy(_._2._1, false)
temp2.take(2)
//Array((10397,(5.431032677813346,ATL)), (13930,(5.4148119418905765,ORD))) 
// get just the airport names
val impAirports =temp2.map(_._2._2)
impAirports.take(4)
//res6: Array[String] = Array(ATL, ORD, DFW, DEN)

Pregel

Many important graph algorithms are iterative algorithms, since properties of vertices depend on properties of their neighbors, which depend on properties of their neighbors. Pregel is an iterative graph processing model, developed at Google, which uses a sequence of iterations of messages passing between vertices in a graph. GraphX implements a Pregel-like bulk-synchronous message-passing API.

With the Pregel implementation in GraphX, vertices can only send messages to neighboring vertices.

The Pregel operator is executed in a series of super steps. In each super step:

  • The vertices receive the sum of their inbound messages from the previous super step
  • They compute a new value for the vertex property
  • They send messages to the neighboring vertices in the next super step

When there are no more messages remaining, the Pregel operator will end the iteration and the final graph is returned.

image03_nstep-relationship-messaging

The code below computes the cheapest airfare using Pregel with the following formula to compute airfare.

50 + distance / 20

// starting vertex
val sourceId: VertexId = 13024
// a graph with edges containing airfare cost calculation
val gg = graph.mapEdges(e => 50.toDouble + e.attr.toDouble/20 )
// initialize graph, all vertices except source have distance infinity
val initialGraph = gg.mapVertices((id, _) => if (id == sourceId) 0.0 else Double.PositiveInfinity)
// call pregel on graph
val sssp = initialGraph.pregel(Double.PositiveInfinity)(
 // Vertex Program
 (id, dist, newDist) => math.min(dist, newDist),
 triplet => {
  // Send Message
  if (triplet.srcAttr + triplet.attr < triplet.dstAttr) {
   Iterator((triplet.dstId, triplet.srcAttr + triplet.attr))
  } else {
   Iterator.empty
  }
 },
 // Merge Message
 (a,b) => math.min(a,b)
) 
// routes , lowest flight cost
println(sssp.edges.take(4).mkString("\n"))
Edge(10135,10397,84.6)
Edge(10135,13930,82.7)
Edge(10140,10397,113.45)
Edge(10140,10821,133.5) 
// routes with airport codes , lowest flight cost
ssp.edges.map{ case ( Edge(org_id, dest_id,price))=> ( (airportMap(org_id), airportMap(dest_id), price)) }.takeOrdered(10)(Ordering.by(_._3))
Array((WRG,PSG,51.55), (PSG,WRG,51.55), (CEC,ACV,52.8), (ACV,CEC,52.8), (ORD,MKE,53.35), (IMT,RHI,53.35), (MKE,ORD,53.35), (RHI,IMT,53.35), (STT,SJU,53.4), (SJU,STT,53.4)) 
// airports , lowest flight cost
println(sssp.vertices.take(4).mkString("\n")) 
(10208,277.79)
(10268,260.7)
(14828,261.65)
(14698,125.25) 
// airport codes , sorted lowest flight cost
sssp.vertices.collect.map(x => (airportMap(x._1), x._2)).sortWith(_._2 < _._2)
res21: Array[(String, Double)] = Array(PDX,62.05), (SFO,65.75), (EUG,117.35)

Want to learn more?

In this blog post, you learned how to get started using Apache Spark GraphX with Scala on the MapR Sandbox. If you have any questions about GraphX, please ask them in the comments section below.

Reference: How to Get Started Using Apache Spark GraphX with Scala from our JCG partner Carol McDonald at the Mapr blog.

 

Subscribe
Notify of
guest

This site uses Akismet to reduce spam. Learn how your comment data is processed.

0 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Back to top button