Big data tools such as Hadoop and Spark allow you to process data at unprecedented scale, but keeping your processing engine fed can be a challenge. Metadata in upstream sources can ‘drift’ due to infrastructure, OS and application changes, causing ETL tools and hand-coded solutions to fail. StreamSets Data Collector (SDC) is an Apache 2.0 licensed open source platform for building big data ingest pipelines that allows you to design, execute and monitor robust data flows. In this session we’ll look at how SDC’s “intent-driven” approach keeps the data flowing, with a particular focus on clustered deployment with Spark and other exciting Spark integrations in the works.

1. Building Data Pipelines with
Spark and StreamSets
Pat Patterson
Community Champion
@metadaddy
pat@streamsets.com

2. Agenda
Data Drift
StreamSets Data Collector
Running Pipelines on Spark Today
Future Spark Integration
Demo

3. Past ETL ETL
Emerging Ingest Analyze
Data Sources Data Stores Data Consumers
The Evolution of Data-in-Motion

4. Data Drift - a Data Engineering Headache
The unpredictable, unannounced and unending mutation of data
characteristics caused by the operation, maintenance and modernization of
the systems that produce the data
Structure
Drift
Semantic
Drift
Infrastructure
Drift

5. SQL on Hadoop (Hive) Y/Y Click Through Rate
80% of analyst time is spent preparing and validating data,
while the remaining 20% is actual data analysis
Example: Data Loss and Corrosion

6. Delayed and
False Insights
Solving Data Drift
Tools
Applications
Data Stores Data Consumers
Poor Data QualityData Drift
Custom code
Fixed-schema
Data Sources
// DIY Custom Code

7. Solving Data Drift
Trusted InsightsData KPIs
Tools
Applications
Data Drift
Intent-Driven
Drift-Handling
Data Stores Data ConsumersData Sources
// DIY Custom Code

8. StreamSets Data Collector
Open source software for the
rapid development and reliably
operation of complex data
flows.
➢ Intent-driven
➢ UI Abstraction
➢ Extensible

9. Handling Drift with Hive
➢ Monitor data structure
➢ Detect schema change
➢ Alter Hive Metadata

10. Running Pipelines on Spark Today
spark-submit
--num-executors ...
--archives ...
--files ...
--jar ...
--class ... ● Container on Spark
● Leverage Kafka RDD
● Scale out for performance

11. SDC on Spark - Connectivity
Sources
• Kafka
Destinations
• HDFS
• HBase
• S3
• Kudu
• MapR DB
• Cassandra
• ElasticSearch
• Kafka
• MapR Streams
• Kinesis
• etc, etc, etc!

12. Future Directions

13. Run Pipelines on Databricks
● Container on Databricks
● Leverage REST API
● Add S3 origin
{
"name":"StreamSets Data Collector",
"new_cluster":{
"num_workers":1
},
"spark_jar_task":{
"main_class_name":"com.streamsets..."
}
}

14. Break Out Spark Processor
● Standalone containers,
Spark processor
● Leverage Spark code
● Custom RDD
● Start local Spark job for each
batch
● Example use cases: running
image classification, sentiment
analysis
Local Mode

15. Spark Processor - Connectivity
Sources
• Kafka
• S3
• MapR Streams
• JDBC
• MongoDB
• Local Filesystem
• Redis
• JMS
• HTTP
• UDP
• etc, etc, etc!
Destinations
• HDFS
• HBase
• S3
• Kudu
• MapR DB
• Cassandra
• ElasticSearch
• Kafka
• MapR Streams
• JDBC
• etc, etc, etc!

16. Deepen Spark Integration
● Container on Spark, Spark processor
● Leverage Spark code
● Custom RDD
● Start Spark job ‘on cluster’ for each pipeline
● Example use cases: training image classification,
sentiment analysis

17. Spark Integration Architecture

18. SDC on Spark - Connectivity Tomorrow
Sources
• Kafka
• S3
• MapR Streams
• JDBC
• MongoDB
• Redis
• JMS
• HTTP
• UDP
• ...any partitionable data source...
Destinations
• HDFS
• HBase
• S3
• Kudu
• MapR DB
• Cassandra
• ElasticSearch
• Kafka
• MapR Streams
• JDBC
• etc, etc, etc!

19. Demo

20. Conclusion
StreamSets Data Collector brings a UI abstraction to Spark
Standalone container + local Spark Processor bring wide
connectivity to Spark code
Spark Container + Spark Processor allow iterative Spark code in
pipelines

21. Resources
Download StreamSets Data Collector
https://streamsets.com/opensource
Contribute Code
https://github.com/streamsets/datacollector
Get Involved
https://streamsets.com/community

22. Thank You!

23. Backup Slides

24. Structure
Drift
Data structures and
formats evolve and change
unexpectedly
Implication:
Data Loss
Data Squandering
Delimited Data
107.3.137.195 fe80::21b:21ff:fe83:90fa
Attribute Format Changes
{
“first“: “jon”
“last“: “smith”
“email“: “jsmith@acme.com”
“add1“: “123 Washington”
“add2“: “”
“city“: “Tucson”
“state“: “AZ”
“zip“: “85756”
}
{
“first“: “jane”
“last“: “smith”
“email“: “jane@earth.net”
“add1“: “456 Fillmore”
“add2“: “Apt 120”
“city“: “Fairfield”
“state“: “VA”
“zip“: “24435-1001”
“phone”: “401-555-1212”
}
Data Structure Evolution
Structure Drift

25. Semantic
Drift
Data semantics change
with evolving applications
Implication:
Data Corrosion
Data Loss
Semantic Drift
24122-52172 00-24122-52172
Account Number Expansion
M134: user {jsmith} read access granted {ac:24122-52172}
M134: user {jsmith} read access granted {ca.ac:24122-52172}
Namespace Qualification
……
…,3588310669797950,$91.41,jcb,K1088-W#9,…
…,6759006011936944,$155.04,switch,A6504-Y#9,…
…,6771111111151415,$37.78,laser,Q9936-T#9,…
…,3585905063294299,$164.48,jcb,S4643-H#9,…
…,5363527828638736,$117.52,mastercard,X3286-P#9,…
…,4903080150282806,$168.03,switch,I9133-W#3,…
……
Outlier / Anomaly Detection

26. Infrastructure
Drift
Physical and Logical
Infrastructure changes
rapidly
Implication:
Poor Agility
Operational Downtime
Data Center 1 Data Center 2 Data Center n
3rd Party Service Provider
App a App k
App q
Cloud
Infrastructure
Infrastructure Drift