GraphX is a new component in Spark for graphs and graph-parallel computation. At a high level, GraphX extends the Spark RDD by introducing a new Graph abstraction: a directed multigraph with properties attached to each vertex and edge.
To support graph computation, GraphX exposes a set of fundamental operators (e.g., subgraph, joinVertices, and aggregateMessages) as well as an optimized variant of the Pregel API from Google.
In addition, GraphX includes a growing collection of graph algorithms and builders to simplify graph analytics tasks.
Example:
Let’s define a simple org chart:
Let’s add one more person, sherry, Jacks’ wife
Here we have data:
The box, indicates the name and title in the org chart:
Sherry, wife of owner
Jack, owner
George, clerk
Mary, sales
The line, indicates the relationship in the org chart:
Jack, owner is boss of George, clerk
Jack, owner is boss of Mary, sales
Sherry, wife of owner is boss of Jack, owner
George, clerk is coworker of Mary, sales
First, I try it on an SQL database. Any database would do. Just use existing HIVE:
First, create 2 tables:
Vertex: stores the Vertex id (long integer), name, title information.
Vertex stores the box or node
Edge: Stores source vertex id, destination vertex id and relationship.
Edge stores the line, or relationship
Create the 2 tables in hive, under database cstu, that was created before
The line, indicates the relationship in the org chart:
Jack, owner is boss of George, clerk
Jack, owner is boss of Mary, sales
Sherry, wife of owner is boss of Jack, owner
George, clerk is coworker of Mary, sales
use cstu;
insert into vertex values (1,'Jack','owner');
insert into vertex values (2, 'George', 'clerk');
insert into vertex values (3, 'Mary', 'Sales');
insert into vertex values (4, 'Shrry', 'wife of owner');
insert into edge values (1,2,'boss');
insert into edge values (1,3,'boss');
insert into edge values (2,3,'coworker');
insert into edge values (4,1,'boss');
Then I construct the SQL query as below:
select * from vertex;
hive> set hive.cli.print.header=true;
hive> select * from vertex;
OK
vertex.id vertex.property_name vertex.property_title
1 Jack owner
2 George clerk
3 Mary Sales
4 Shrry wife of owner
select * from edge;
hive> set hive.cli.print.header=true;
hive> select * from edge;
OK
edge.src_id edge.dest_id edge.relationship
1 2 boss
1 3 boss
2 3 coworker
4 1 boss
Here is the Graph query to describe the org chart.
with x as (SELECT e.src_id, e.dest_id, e.relationship,
src.property_name src_name,
src.property_title src_title,
dst.property_name dest_name,
dst.property_title dest_title
FROM edge AS e LEFT JOIN vertex AS src ON e.src_id = src.id
LEFT JOIN vertex AS dst ON e.dest_id = dst.id)
select concat_ws('',src_name, ', ', src_title , ', is '
, relationship , ' of ' , dest_name , ', '
, dest_title
) from x;
Run it in hive
(spark) [hadoop@master ~]$ hive
which: no hbase in (/opt/spark/bin:/opt/hadoop/hive/bin:/opt/hadoop/anaconda3/envs/spark/bin:/opt/hadoop/anaconda3/condabin:/usr/local/bin:/usr/bin:/usr/local/sbin:/usr/sbin:/opt/hadoop/.local/bin:/opt/hadoop/bin:/usr/java/default/bin:/opt/hadoop/sbin:/opt/hadoop/bin)
SLF4J: Class path contains multiple SLF4J bindings.
SLF4J: Found binding in [jar:file:/opt/hadoop/hive/lib/log4j-slf4j-impl-2.4.1.jar!/org/slf4j/impl/StaticLoggerBinder.class]
SLF4J: Found binding in [jar:file:/opt/hadoop/share/hadoop/common/lib/slf4j-log4j12-1.7.10.jar!/org/slf4j/impl/StaticLoggerBinder.class]
SLF4J: See http://www.slf4j.org/codes.html#multiple_bindings for an explanation.
SLF4J: Actual binding is of type [org.apache.logging.slf4j.Log4jLoggerFactory]
Logging initialized using configuration in jar:file:/opt/hadoop/hive/lib/hive-common-2.1.0.jar!/hive-log4j2.properties Async: true
Hive-on-MR is deprecated in Hive 2 and may not be available in the future versions. Consider using a different execution engine (i.e. spark, tez) or using Hive 1.X releases.
hive> use cstu;
hive> with x as (SELECT e.src_id, e.dest_id, e.relationship,
> src.property_name src_name,
> src.property_title src_title,
> dst.property_name dest_name,
> dst.property_title dest_title
> FROM edge AS e LEFT JOIN vertex AS src ON e.src_id = src.id
> LEFT JOIN vertex AS dst ON e.dest_id = dst.id)
> select concat_ws('',src_name, ', ', src_title , ', is '
> , relationship , ' of ' , dest_name , ', '
> , dest_title
> ) from x;
Here is the result of query, compare with the org chart, it that right?
Jack, owner, is boss of George, clerk
Jack, owner, is boss of Mary, Sales
George, clerk, is coworker of Mary, Sales
Shrry, wife of owner, is boss of Jack, owner
Now switch to Spark GraphX library with Scala:
import org.apache.spark._
import org.apache.spark.graphx._
import org.apache.spark.rdd.RDD
import org.apache.log4j._
import org.apache.spark.sql._
import org.apache.spark.graphx.{Graph, VertexRDD}
import org.apache.spark.graphx.util.GraphGenerators
Logger.getLogger("org").setLevel(Level.ERROR)
val spark = SparkSession
.builder
.appName("graphx")
.master("local[*]")
.config("spark.sql.warehouse.dir", "file:///tmp")
.getOrCreate()
val sc=spark.sparkContext
val users: RDD[(VertexId, (String, String))] =
sc.parallelize(Array((1L, ("jack", "owner")), (2L, ("george", "clerk")),
(3L, ("mary", "sales")), (4L, ("sherry", "owner wife"))))
val relationships: RDD[Edge[String]] =
sc.parallelize(Array(Edge(1L, 2L, "boss"), Edge(1L, 3L, "boss"),
Edge(2L, 3L, "coworker"), Edge(4L, 1L, "boss")))
val defaultUser = ("", "Missing")
val graph = Graph(users, relationships, defaultUser)
/*
The EdgeTriplet class extends the Edge class by adding the srcAttr and dstAttr members
which contain the source and destination properties respectively.
We can use the triplet view of a graph to render a collection of strings describing relationships between users.
*/
val facts: RDD[String] =
graph.triplets.map(triplet =>
triplet.srcAttr._1 + ", "+ triplet.srcAttr._2 + " is the " + triplet.attr + " of " + triplet.dstAttr._1+", "+triplet.dstAttr._2)
facts.collect.foreach(println(_))
Run the above Scala code, output below:
jack, owner is the boss of george, clerk
jack, owner is the boss of mary, sales
george, clerk is the coworker of mary, sales
sherry, owner wife is the boss of jack, owner
If this is a complex charts, Spark Graphx is the perfect platform to compute the inter-relationships because of its distributed, in memory computing nature.