This document discusses a new approach to building scalable data processing systems using streaming analytics with Spark, Kafka, Cassandra, and Akka. It proposes moving away from architectures like Lambda and ETL that require duplicating data and logic. The new approach leverages Spark Streaming for a unified batch and stream processing runtime, Apache Kafka for scalable messaging, Apache Cassandra for distributed storage, and Akka for building fault tolerant distributed applications. This allows building real-time streaming applications that can join streaming and historical data with simplified architectures that remove the need for duplicating data extraction and loading.

1. Streaming Analytics with Spark,
Kafka, Cassandra, and Akka
Helena Edelson
VP of Product Engineering @Tuplejump

2. • Committer / Contributor: Akka, FiloDB, Spark Cassandra
Connector, Spring Integration
• VP of Product Engineering @Tuplejump
• Previously: Sr Cloud Engineer / Architect at VMware,
CrowdStrike, DataStax and SpringSource
Who
@helenaedelson
github.com/helena

3. Tuplejump
Tuplejump Data Blender combines sophisticated data collection
with machine learning and analytics, to understand the intention of
the analyst, without disrupting workflow.
• Ingest streaming and static data from disparate data sources
• Combine them into a unified, holistic view
• Easily enable fast, flexible and advanced data analysis
3

4. Tuplejump Open Source
github.com/tuplejump
• FiloDB - distributed, versioned, columnar analytical db for modern
streaming workloads
• Calliope - the first Spark-Cassandra integration
• Stargate - Lucene indexer for Cassandra
• SnackFS - HDFS-compatible file system for Cassandra
4

5. What Will We Talk About
• The Problem Domain
• Example Use Case
• Rethinking Architecture
– We don't have to look far to look back
– Streaming
– Revisiting the goal and the stack
– Simplification

6. THE PROBLEM DOMAIN
Delivering Meaning From A Flood Of Data
6

7. The Problem Domain
Need to build scalable, fault tolerant, distributed data
processing systems that can handle massive amounts of
data from disparate sources, with different data structures.
7

8. Translation
How to build adaptable, elegant systems
for complex analytics and learning tasks
to run as large-scale clustered dataflows
8

9. How Much Data
Yottabyte = quadrillion gigabytes or septillion
bytes
9
We all have a lot of data
• Terabytes
• Petabytes...
https://en.wikipedia.org/wiki/Yottabyte

10. Delivering Meaning
• Deliver meaning in sec/sub-sec latency
• Disparate data sources & schemas
• Billions of events per second
• High-latency batch processing
• Low-latency stream processing
• Aggregation of historical from the stream

11. While We Monitor, Predict & Proactively Handle
• Massive event spikes
• Bursty traffic
• Fast producers / slow consumers
• Network partitioning & Out of sync systems
• DC down
• Wait, we've DDOS'd ourselves from fast streams?
• Autoscale issues
– When we scale down VMs how do we not lose data?

12. And stay within our
AWS / Rackspace budget

13. EXAMPLE CASE:
CYBER SECURITY
Hunting The Hunter
13

14. 14
• Track activities of international threat actor groups,
nation-state, criminal or hactivist
• Intrusion attempts
• Actual breaches
• Profile adversary activity
• Analysis to understand their motives, anticipate actions
and prevent damage
Adversary Profiling & Hunting

15. 15
• Machine events
• Endpoint intrusion detection
• Anomalies/indicators of attack or compromise
• Machine learning
• Training models based on patterns from historical data
• Predict potential threats
• profiling for adversary Identification
•
Stream Processing

16. Data Requirements & Description
• Streaming event data
• Log messages
• User activity records
• System ops & metrics data
• Disparate data sources
• Wildly differing data structures
16

17. Massive Amounts Of Data
17
• One machine can generate 2+ TB per day
• Tracking millions of devices
• 1 million writes per second - bursty
• High % writes, lower % reads
• TTL

18. RETHINKING
ARCHITECTURE
18

19. WE DON'T HAVE TO LOOK
FAR TO LOOK BACK
19
Rethinking Architecture

20. 20
Most batch analytics flow from
several years ago looked like...

21. STREAMING & DATA SCIENCE
21
Rethinking Architecture

22. Streaming
I need fast access to historical data on the fly for
predictive modeling with real time data from the stream.
22

23. Not A Stream, A Flood
• Data emitters
• Netflix: 1 - 2 million events per second at peak
• 750 billion events per day
• LinkedIn: > 500 billion events per day
• Data ingesters
• Netflix: 50 - 100 billion events per day
• LinkedIn: 2.5 trillion events per day
• 1 Petabyte of streaming data
23

24. Which Translates To
• Do it fast
• Do it cheap
• Do it at scale
24

25. Challenges
• Code changes at runtime
• Distributed Data Consistency
• Ordering guarantees
• Complex compute algorithms
25

26. Oh, and don't lose data
26

27. Strategies
• Partition For Scale & Data Locality
• Replicate For Resiliency
• Share Nothing
• Fault Tolerance
• Asynchrony
• Async Message Passing
• Memory Management
27
• Data lineage and reprocessing in
runtime
• Parallelism
• Elastically Scale
• Isolation
• Location Transparency

28. AND THEN WE GREEKED OUT
28
Rethinking Architecture

29. Lambda Architecture
A data-processing architecture designed to handle massive
quantities of data by taking advantage of both batch and
stream processing methods.
29

30. Lambda Architecture
A data-processing architecture designed to handle massive
quantities of data by taking advantage of both batch and
stream processing methods.
• An approach
• Coined by Nathan Marz
• This was a huge stride forward
30

31. 31
https://www.mapr.com/developercentral/lambda-architecture

33. Implementing Is Hard
33
• Real-time pipeline backed by KV store for updates
• Many moving parts - KV store, real time, batch
• Running similar code in two places
• Still ingesting data to Parquet/HDFS
• Reconcile queries against two different places

34. Performance Tuning & Monitoring:
on so many systems
34
Also hard

35. Lambda Architecture
An immutable sequence of records is captured and fed
into a batch system and a stream processing
system in parallel.
35

36. WAIT, DUAL SYSTEMS?
36
Challenge Assumptions

37. Which Translates To
• Performing analytical computations & queries in dual
systems
• Implementing transformation logic twice
• Duplicate Code
• Spaghetti Architecture for Data Flows
• One Busy Network
37

38. Why Dual Systems?
• Why is a separate batch system needed?
• Why support code, machines and running services of
two analytics systems?
38
Counter productive on some level?

39. YES
39
• A unified system for streaming and batch
• Real-time processing and reprocessing
• Code changes
• Fault tolerance
http://radar.oreilly.com/2014/07/questioning-the-lambda-
architecture.html - Jay Kreps

40. ANOTHER ASSUMPTION:
ETL
40
Challenge Assumptions

41. Extract, Transform, Load (ETL)
41
"Designing and maintaining the ETL process is often
considered one of the most difficult and resource-
intensive portions of a data warehouse project."
http://docs.oracle.com/cd/B19306_01/server.102/b14223/ettover.htm

42. Extract, Transform, Load (ETL)
42
ETL involves
• Extraction of data from one system into another
• Transforming it
• Loading it into another system

43. Extract, Transform, Load (ETL)
"Designing and maintaining the ETL process is often
considered one of the most difficult and resource-
intensive portions of a data warehouse project."
http://docs.oracle.com/cd/B19306_01/server.102/b14223/ettover.htm
43
Also unnecessarily redundant and often typeless

44. ETL
44
• Each ETL step can introduce errors and risk
• Can duplicate data after failover
• Tools can cost millions of dollars
• Decreases throughput
• Increased complexity

45. ETL
• Writing intermediary files
• Parsing and re-parsing plain text
45

46. And let's duplicate the pattern
over all our DataCenters
46

47. 47
These are not the solutions you're looking for

48. REVISITING THE GOAL
& THE STACK
48

49. Removing The 'E' in ETL
Thanks to technologies like Avro and Protobuf we don’t need the
“E” in ETL. Instead of text dumps that you need to parse over
multiple systems:
Scala & Avro (e.g.)
• Can work with binary data that remains strongly typed
• A return to strong typing in the big data ecosystem
49

50. Removing The 'L' in ETL
If data collection is backed by a distributed messaging
system (e.g. Kafka) you can do real-time fanout of the
ingested data to all consumers. No need to batch "load".
• From there each consumer can do their own transformations
50

51. #NoMoreGreekLetterArchitectures
51

52. NoETL
52

53. Strategy Technologies
Scalable Infrastructure / Elastic Spark, Cassandra, Kafka
Partition For Scale, Network Topology Aware Cassandra, Spark, Kafka, Akka Cluster
Replicate For Resiliency Spark,Cassandra, Akka Cluster all hash the node ring
Share Nothing, Masterless Cassandra, Akka Cluster both Dynamo style
Fault Tolerance / No Single Point of Failure Spark, Cassandra, Kafka
Replay From Any Point Of Failure Spark, Cassandra, Kafka, Akka + Akka Persistence
Failure Detection Cassandra, Spark, Akka, Kafka
Consensus & Gossip Cassandra & Akka Cluster
Parallelism Spark, Cassandra, Kafka, Akka
Asynchronous Data Passing Kafka, Akka, Spark
Fast, Low Latency, Data Locality Cassandra, Spark, Kafka
Location Transparency Akka, Spark, Cassandra, Kafka
My Nerdy Chart
53

54. SMACK
• Scala/Spark
• Mesos
• Akka
• Cassandra
• Kafka
54

55. Spark Streaming
55

56. Spark Streaming
• One runtime for streaming and batch processing
• Join streaming and static data sets
• No code duplication
• Easy, flexible data ingestion from disparate sources to
disparate sinks
• Easy to reconcile queries against multiple sources
• Easy integration of KV durable storage
56

57. How do I merge historical data with data
in the stream?
57

58. Join Streams With Static Data
val ssc = new StreamingContext(conf, Milliseconds(500))
ssc.checkpoint("checkpoint")
val staticData: RDD[(Int,String)] =
ssc.sparkContext.textFile("whyAreWeParsingFiles.txt").flatMap(func)
val stream: DStream[(Int,String)] =
KafkaUtils.createStream(ssc, zkQuorum, group, Map(topic -> n))
.transform { events => events.join(staticData))
.saveToCassandra(keyspace,table)
ssc.start()
58

59. Training
Data
Feature
Extraction
Model
Training
Model
Testing
Test Data
Your Data Extract Data To Analyze
Train your model to predict
Spark MLLib
59

60. Spark Streaming & ML
60
val context = new StreamingContext(conf, Milliseconds(500))
val model = KMeans.train(dataset, ...) // learn offline
val stream = KafkaUtils
.createStream(ssc, zkQuorum, group,..)
.map(event => model.predict(event.feature))

61. Apache Mesos
Open-source cluster manager developed at UC Berkeley.
Abstracts CPU, memory, storage, and other compute resources
away from machines (physical or virtual), enabling fault-tolerant
and elastic distributed systems to easily be built and run
effectively.
61

62. Akka
High performance concurrency framework for Scala and
Java
• Fault Tolerance
• Asynchronous messaging and data processing
• Parallelization
• Location Transparency
• Local / Remote Routing
• Akka: Cluster / Persistence / Streams
62

63. Akka Actors
A distribution and concurrency abstraction
• Compute Isolation
• Behavioral Context Switching
• No Exposed Internal State
• Event-based messaging
• Easy parallelism
• Configurable fault tolerance
63

64. 64
Akka Actor Hierarchy
http://www.slideshare.net/jboner/building-reactive-applications-with-akka-in-scala

65. import akka.actor._
class NodeGuardianActor(args...) extends Actor with SupervisorStrategy {
val temperature = context.actorOf(
Props(new TemperatureActor(args)), "temperature")
val precipitation = context.actorOf(
Props(new PrecipitationActor(args)), "precipitation")
override def preStart(): Unit = { /* lifecycle hook: init */ }
def receive : Actor.Receive = {
case Initialized => context become initialized
}
def initialized : Actor.Receive = {
case e: SomeEvent => someFunc(e)
case e: OtherEvent => otherFunc(e)
}
}
65

66. Apache Cassandra
• Extremely Fast
• Extremely Scalable
• Multi-Region / Multi-Datacenter
• Always On
• No single point of failure
• Survive regional outages
• Easy to operate
• Automatic & configurable replication 66

67. Apache Cassandra
• Very flexible data modeling (collections, user defined
types) and changeable over time
• Perfect for ingestion of real time / machine data
• Huge community
67

68. Spark Cassandra Connector
• NOSQL JOINS!
• Write & Read data between Spark and Cassandra
• Compatible with Spark 1.4
• Handles Data Locality for Speed
• Implicit type conversions
• Server-Side Filtering - SELECT, WHERE, etc.
• Natural Timeseries Integration
68
http://github.com/datastax/spark-cassandra-connector

69. KillrWeather
69
http://github.com/killrweather/killrweather
A reference application showing how to easily integrate streaming and
batch data processing with Apache Spark Streaming, Apache
Cassandra, Apache Kafka and Akka for fast, streaming computations
on time series data in asynchronous event-driven environments.
http://github.com/databricks/reference-apps/tree/master/timeseries/scala/timeseries-weather/src/main/scala/com/
databricks/apps/weather

70. 70
• High Throughput Distributed Messaging
• Decouples Data Pipelines
• Handles Massive Data Load
• Support Massive Number of Consumers
• Distribution & partitioning across cluster nodes
• Automatic recovery from broker failures

71. Spark Streaming & Kafka
val context = new StreamingContext(conf, Seconds(1))
val wordCount = KafkaUtils.createStream(context, ...)
.flatMap(_.split(" "))
.map(x => (x, 1))
.reduceByKey(_ + _)
wordCount.saveToCassandra(ks,table)
context.start() // start receiving and computing
71

72. 72
class KafkaStreamingActor(params: Map[String, String], ssc: StreamingContext)
extends AggregationActor(settings: Settings) {
import settings._
val kafkaStream = KafkaUtils.createStream[String, String, StringDecoder, StringDecoder](
ssc, params, Map(KafkaTopicRaw -> 1), StorageLevel.DISK_ONLY_2)
.map(_._2.split(","))
.map(RawWeatherData(_))
kafkaStream.saveToCassandra(CassandraKeyspace, CassandraTableRaw)
/** RawWeatherData: wsid, year, month, day, oneHourPrecip */
kafkaStream.map(hour => (hour.wsid, hour.year, hour.month, hour.day, hour.oneHourPrecip))
.saveToCassandra(CassandraKeyspace, CassandraTableDailyPrecip)
/** Now the [[StreamingContext]] can be started. */
context.parent ! OutputStreamInitialized
def receive : Actor.Receive = {…}
}
Gets the partition key: Data Locality
Spark C* Connector feeds this to Spark
Cassandra Counter column in our schema,
no expensive `reduceByKey` needed. Simply
let C* do it: not expensive and fast.

73. 73
/** For a given weather station, calculates annual cumulative precip - or year to date. */
class PrecipitationActor(ssc: StreamingContext, settings: WeatherSettings) extends AggregationActor {
def receive : Actor.Receive = {
case GetPrecipitation(wsid, year) => cumulative(wsid, year, sender)
case GetTopKPrecipitation(wsid, year, k) => topK(wsid, year, k, sender)
}
/** Computes annual aggregation.Precipitation values are 1 hour deltas from the previous. */
def cumulative(wsid: String, year: Int, requester: ActorRef): Unit =
ssc.cassandraTable[Double](keyspace, dailytable)
.select("precipitation")
.where("wsid = ? AND year = ?", wsid, year)
.collectAsync()
.map(AnnualPrecipitation(_, wsid, year)) pipeTo requester
/** Returns the 10 highest temps for any station in the `year`. */
def topK(wsid: String, year: Int, k: Int, requester: ActorRef): Unit = {
val toTopK = (aggregate: Seq[Double]) => TopKPrecipitation(wsid, year,
ssc.sparkContext.parallelize(aggregate).top(k).toSeq)
ssc.cassandraTable[Double](keyspace, dailytable)
.select("precipitation")
.where("wsid = ? AND year = ?", wsid, year)
.collectAsync().map(toTopK) pipeTo requester
}
}

74. A New Approach
• One Runtime: streaming, scheduled
• Simplified architecture
• Allows us to
• Write different types of applications
• Write more type safe code
• Write more reusable code
74

75. Need daily analytics aggregate reports? Do it in the stream, save
results in Cassandra for easy reporting as needed - with data
locality not offered by S3.

76. FiloDB
Distributed, columnar database designed to run very fast
analytical queries
• Ingest streaming data from many streaming sources
• Row-level, column-level operations and built in versioning
offer greater flexibility than file-based technologies
• Currently based on Apache Cassandra & Spark
• github.com/tuplejump/FiloDB
76

77. FiloDB
• Breakthrough performance levels for analytical queries
• Performance comparable to Parquet
• One to two orders of magnitude faster than Spark on
Cassandra 2.x
• Versioned - critical for reprocessing logic/code changes
• Can simplify your infrastructure dramatically
• Queries run in parallel in Spark for scale-out ad-hoc analysis
• Space-saving techniques
77

78. WRAPPING UP
78

79. Architectyr?
79
"This is a giant mess"
- Going Real-time - Data Collection and Stream Processing with Apache Kafka, Jay Kreps

80. 80
Simplified

81. 81

82. 82
www.tuplejump.com
info@tuplejump.com@tuplejump

83. 83
@helenaedelson
github.com/helena
slideshare.net/helenaedelson
THANK YOU!

84. I'm speaking at QCon SF on the broader
topic of Streaming at Scale
http://qconsf.com/sf2015/track/streaming-data-scale
84