Graph Analytics

DSE Graph can be used in conjunction with Apache Spark embedded in DSE to perform Online Analytical Processing (OLAP) queries on graph datasets. OLAP can be enabled for graph queries by simply setting the graph_source graph option to ‘a’ or using the :default_graph_analytics execution profile. Graph OLAP queries will always be routed to the Spark master.

Background

Given

a running dse cluster with graph and spark enabled

And

an existing graph called “user_connections_spark” with schema:

schema.propertyKey('name').Text().ifNotExists().create();
schema.propertyKey('age').Int().ifNotExists().create();
schema.propertyKey('lang').Text().ifNotExists().create();
schema.propertyKey('weight').Float().ifNotExists().create();
schema.vertexLabel('person').properties('name', 'age').ifNotExists().create();
schema.vertexLabel('software').properties('name', 'lang').ifNotExists().create();

Vertex marko = graph.addVertex(label, 'person', 'name', 'marko', 'age', 29);
Vertex vadas = graph.addVertex(label, 'person', 'name', 'vadas', 'age', 27);
Vertex lop = graph.addVertex(label, 'software', 'name', 'lop', 'lang', 'java');
Vertex josh = graph.addVertex(label, 'person', 'name', 'josh', 'age', 32);
Vertex ripple = graph.addVertex(label, 'software', 'name', 'ripple', 'lang', 'java');
Vertex peter = graph.addVertex(label, 'person', 'name', 'peter', 'age', 35);

Running an OLAP graph query

Given

the following example:

require 'dse'

cluster = Dse.cluster(graph_name: 'user_connections_spark')
session = cluster.connect

results = session.execute_graph('g.V().count()', execution_profile: :default_graph_analytics)
puts "Result: #{results.first.value}"
puts "The spark master was: #{results.execution_info.hosts.last.ip}"

When

it is executed

Then

its output should contain:

Result: 6
The spark master was: 127.0.0.1