Stratosphere System Overview Big Data Beers Berlin. 20.11.2013

Stratosphere: System Overview

Robert [email protected]

Twitter: @rmetzger_

Big Data Beers Meetup, Nov. 19th, 2013

Stratosphere

… is a distributed data processing engine

… automatically handles parallelization

… brings database technology to the world of big data

● Extends MapReduce with more operators

● Support for advanced data flow graphs

● Compiler/Optimizer, Java/Scala Interface, YARN

Overview

map crossjoin cogroup

Known from Hadoop New in Stratosphere

reduce

Known from Hadoop New in Stratosphere

M R

M R

J

M

R R

Local Files HDFS S3 ...Storage

YARN Direct EC2Cluster

Manager

Stratosphere Runtime

Stratosphere Optimizer

Java API

Scala API Meteor ...

Stratosphere System Stack

Hadoop MR

Hive

Stratosphere in a Cluster

● Operators are executed over the whole cluster

● Side by side with Hadoop● Scales by adding more

nodes● Support for YARN is in

development● We have a LocalExecutor

Cluster Node

JobSubmission

JobManagerResource Mgmt

CompilerWeb Interface

Master Node

TaskManager

DataNode

TaskManager

DataNode

TaskManager

4 Worker Nodes

DataNode

TaskManager

DataNodeStratosphere

Hadoop

Legend:

4. Scala Interface

1. Data Flows

3. Iterations

2. Optimizer

Data Flows: Execution Models

One of many possible data flows in Stratosphere

M R

J

M

R

M RApache Hadoop MR is limited to one data flow

Complex Data Flows in Hadoop

M R

M R

M

M R

R

J

M

RFiltering

Joining

Grouping

Grouping

Data Flows: Lessons Learned

1. Most tasks do not fit the MapReduce model2. Very expensive

○ Always go to disk and HDFS

3. Tedious to implement○ Custom data types and file formats between jobs

That’s why higher level abstractions for MR exist.

R

J

M

R

● Data flow graphs are supported natively● Stratosphere only writes to disk if necessary,

otherwise in-memory

Advanced Data Flows in Stratosphere

Skeleton of a Stratosphere Program

● Input: text file, JDBC source, CSV, etc. ● Transformations

○ map, reduce, join, iterate etc.

● Output: to file etc.● Data Types

○ PactRecord: Tuples with n fields.

○ custom data types for vectors, images, audio (we only expect serialization and compare)

2

Data Flows: Code Example

FileDataSource customers = new FileDataSource(TextInputFormat.class, customersPath);

FileDataSource orders = new FileDataSource(TextInputFormat.class, ordersPath);

MapContract ordersFiltered = MapContract.builder(FilterOrders.class)

.input(orders).build();

ReduceContract groupedCustomers = ReduceContract.builder(GroupCustomers.class)

.input(customers)

.keyField(PactInteger.class, 0).build();

MatchContract joined = MatchContract.builder(JoinOnCustomerid.class, PactInteger.class, 0,

0)

.input1(ordersFiltered)

.input2(groupedCustomers).build();

ReduceContract orderBy = ReduceContract.builder(MaxSum.class)

.input(joined)

.keyField(PactInteger.class, 0).build();

FileDataSink sink = new FileDataSink(RecordOutputFormat.class, outputPath, orderBy);

R

J

M

R

Filter Mapper

Define group key

Map Stub and PactRecord by Example

public class FilterOrders extends MapStub {

@Override

public void map(PactRecord order, Collector<PactRecord> out)

throws Exception {

PactString date = order.getField(Orders.DATE_IDX, PactString.class);

if (date.getValue().equals("11.20.2013")) {

out.collect(order);

}

}

}

MapContract ordersFiltered = MapContract.builder(FilterOrders.class)

.input(orders).build();

4. Scala Interface

1. Data Flows

3. Iterations

2. Optimizer

Joins in Hadoop

● Which strategy to choose?● How to configure it

Lessons Learned:

● Joins do not naturally fit MapReduce● Very time consuming to implement● Hand optimization necessary

Source: Sebastian Schelter, TU Berlin

Map (Broadcast) Join Reduce (Repartition) Join

Joins with Stratosphere

● Natively implemented into the system● Optimizer decides join strategy:

○ Sort-merge-join○ Hybrid Hash Join○ Data Shipping Strategy

● Hybrid Hash Join starts in-memory and gracefully degrades to disk

Optimizer Magic

Recap example job:

● We require a grouped input for the reducer (sorting or hashing)

● Optimizer chooses Sort-Merge-Join → no sorting for reduce

R

J

M

RFiltering

Joining

Grouping

Grouping

Stratosphere Optimizer

● Cost-based optimizer○ Enumerate different execution plans○ Choose the cheapest one

● Optimizer collects statistics○ Size of input and output

● Operators (Map, Reduce, Join) tell how they modify fields

● In-memory chaining of operators● Memory Distribution

⇒ Focus on your application logic rather than parallel execution.

4. Scala Interface

1. Data Flows

3. Iterations

2. Optimizer

Algorithms that need iterations

● K-Means● Gradient descent● Page-Rank● Logistic Regression● Path algorithms on graphs● Graph communities / dense sub-components● Inference (belief propagation)

● Many algorithms loop over the data○ Machine learning: iteratively refine the model○ Graph processing: propagate information hop by hop

Why Iterations?

Initial Input

1 2

3 4

5

6 7

1st Iteration

1 1

2 2

5

5 5

2nd Iteration

1 1

1 1

5

5 5

Example: Connected Components

● Loop is outside the system○ Hard to program○ Very poor performance

Iterations in Hadoop

M R M R M R

1st Iteration

Usually each iteration is more than a single map and reduce!

n-th Iteration

Driver

Spawn 1st Iteratio

n

2nd Iteration ...

Spaw

n 2n

d It

erat

ion Spawn n-th Iteration

● Loop is inside the system○ Easy to program○ Huge performance gains

Iterations in Stratosphere

M

C

M

R R M

Iterate

4. Scala Interface

1. Data Flows

3. Iterations

2. Optimizer

● Functional object oriented programming language● ScaLa = Scalable Language● Very productive (few LOC)● Feels like a scripting language● No more UDFs● Easy to integrate● Runs in JVM, is compatible to regular Java classes● Basis for developing embedded domain specific

languages (DSL)

Do more, write less!class Person(val firstName: String, val lastName: String)

public class Person {

private final String firstName;

private final String lastName;

public Person(String firstName, String lastName) {

this.firstName = firstName;

this.lastName = lastName;

}

public String getFirstName() {

return firstName;

}

public String getLastName() {

return lastName;

}

}

Let the code speak

val input = TextFile(textInput)

val words = input.flatMap { line => line.split(" ") }

val counts = words

.groupBy { word => word }

.count()

val output = counts.write(wordsOutput, CsvOutputFormat())

val plan = new ScalaPlan(Seq(output))

Example in Scala

FileDataSource customers = new FileDataSource(TextInputFormat.class, customersPath);

FileDataSource orders = new FileDataSource(TextInputFormat.class, ordersPath);

MapContract ordersFiltered = MapContract.builder(FilterOrders.class).input(orders).build();

ReduceContract groupedCustomers = ReduceContract.builder(GroupCustomers.class)

.input(customers)

.keyField(PactInteger.class, 0)

.build();

MatchContract joined = MatchContract.builder(JoinOnCustomerid.class,PactInteger.class, 0,0)

.input1(ordersFiltered).input2(groupedCustomers).build();

ReduceContract orderBy = ReduceContract.builder(MaxSum.class)

.input(joined)

.keyField(PactInteger.class, 0)

.build();

FileDataSink sink = new FileDataSink(RecordOutputFormat.class, outputPath, orderBy, "output: word counts");

val customers = DataSource(customersPath, CsvInputFormat[Customer])

val orders = DataSource(ordersPath, CsvInputFormat[Order])

val ordersFiltered = orders filter { order => order.date.equals("11.20.2013")}

val groupedCustomers = customers groupBy { cust => cust.zip} reduceGroup {grp => (grp.buffered.head.zip,

grp.maxBy{_.total})}

val joined = ordersFiltered .join(groupedCustomers) .where {ord => ord.c_id}

.isEqualTo {cust => cust._1} .map { (orders, cust) => cust}

val max = joined groupBy { cust => cust.category_id} reduceGroup {_.maxBy{_.sum}}

val output = counts.write(wordsOutput, DelimitedOutputFormat(formatOutput.tupled))

val plan = new ScalaPlan(Seq(output), "BDB Example")

R

J

M

R

Stratosphere: Database inspired Big Data Analytics

Summary: Feature Matrix

Map Reduce Stratosphere

Operators

● Map● Reduce

● Map● Reduce (multiple sort keys)● Cross● Join● CoGroup● Union● Iterate, Iterate Delta

Composition Only MapReduce Arbitrary Data flows

Data Exchange Batch through diskPipelined, in-memory

(automatic spilling to disk)

Stratosphere is the next-generation open source Big Data Analytics Platform.

Quickstart: http://stratosphere.eu/quickstart

Website: http://stratosphere.eu

GitHub: https://github.com/stratosphere

Mailing List:https://groups.google.com/d/forum/stratosphere-dev

Twitter: @stratosphere_eu

Get In Touch