Scc2012 Scala

cipi

Centro di Ricerca sull’Ingegneria delle Piattaforme Informatiche

Event-Driven Mashup Orchestration with Scala

Michele Stecca, Martino Fornasa, Nicholas Dall’Armellina, and Massimo Maresca

Follow me on Twitter: @steccamiE-mail: [email protected]

Computer Platform Research Center (CIPI)

University of Padova & Genova (Italy)

Honolulu, HI, June 28th, 2012

cipi


2

Agenda

1. Introduction

2. Mashup Execution Platform (MEP) Design

3. The Scala programming language

4. A Scala-based MEP

5. Performance evaluation

6. Conclusions

cipi


…about the ‘Mashup’ concept

Data Mashup: combining different data sources (e.g., Yahoo!Pipes)

Presentation Mashup: combining different gadgets/widgets in a web page (e.g., iGoogle)

Process/Functional Mashup: combining different services/Web APIs (e.g., JackBe Presto but also IFTTT)

1. Introduction (1/5)

cipi


4

1. Introduction (2/5)

Reference Scenario Availability of contents and services through technologies

typical of the Web 2.0 philosophy such as RSS Feed, Atom, REST-WS, SOAP-WS, etc. See programmableweb.com for a list of more than 6000 Web APIs

Internet/Web of Things (need for event management)

Availability of tools for the rapid development of convergent Composite Services (a.k.a., Mashups) that combine different resources such as Yahoo Pipes!, JackBe Presto, etc.

cipi


The Mashup Creation Platform1. Introduction (3/5)

cipi


The Mashup Creation Platform1. Introduction (4/5)

Check if RSS item contains “Italy defeated Germany”

Event Date: July 1st, 2012Event Title: Final Match

TweetText = RSS_Eurosport.title

cipi


Mashup Repository

Mashup Execution Platform(Server-side platform)

Mashup Creation Platform

1. Introduction (5/5)System overview

cipi


8

2. The MEP Design (1/2)

We already implemented a Java-based MEP* whose high level architecture is depicted in the figure

*Stecca M., Maresca M., “An Architecture for a Mashup Container in Virtualized Environments”, Cloud, pp.386-393, 2010 IEEE 3rd International Conference on Cloud Computing, Miami, Florida, 2010

SP=Service Proxy (i.e., a wrapper for Web APIs)

cipi


9

2. The MEP Design (2/2)

The MEP has been designed as a Server-side platform “in-the-cloud” (reasons: security, performance optimization, always-on, etc.)

It is important to chose a suitable concurrency model because of the huge number of Mashup executions taking place concurrently

Two approaches: Thread-based (but difficult to optimize the resource usage) Event-driven (but difficult to develop)

How can we improve the existing Java-based MEP performance without re-developing the system from the scratch?

Possible answer: by using the Scala programming language

cipi


10

3. The Scala programming language (1/4) Developed at the EPFL Lausanne,

Switzerland It merges the Functional Programming (FP)

and the OOP paradigms It is executed by a JVM Fully compatibility

with legacy Java libraries Concurrency model: actors (e.g., Erlang)

It is suitable to exploit the multi-core processors An actor is a “computational unit” lighter than a

system thread Actors communicate among each others

according to the message passing paradigm DestinationActor ! Message

cipi


11

3. The Scala programming language (2/4)

Here is the typical Scala Actor’s body (pseudocode)

Loop{ onMessage{ Case Message1=> Process Message1;

Case Message2=> Process Message2;

… }

}

cipi


12


Case 1: onMessage=receive primitive

The concurrency model falls back to the Thread model

cipi


13


Case 2: onMessage=react primitive

The concurrency model is event driven (a.k.a., threadless model)

cipi


14

4. A Scala-based MEP (1/4) The Java-based MEP has been developed in

modular way (i.e., clear separation between the Business Logic and the Concurrency logic)

The idea is to exploit at the same time: The Java-based MEP implementation The efficient Scala concurrency model

based on actors What have we done?

We mapped the Orchestrator component and each SP into Scala actors

We defined the MessageDo and the MessageEvent Scala messages

cipi


15

4. A Scala-based MEP (2/4)

cipi


16

4. A Scala-based MEP (3/4) The reference example OAI=Orchestrator Actor Instance

cipi


17

4. A Scala-based MEP (4/4)The Orchestrator component in Scala (code)

cipi


18

5. Performance Evaluation (1/2)The testbed Two nodes (QuadCore, 4GB RAM, Ubuntu OS)

Node 1: traffic injector Node 2: the Scala-based MEP

Deployment of 2 fake SPs and 1 fake Mashup Definition of two metrics:

Mashup Execution Time (MET) #Threads

cipi


19

5. Performance Evaluation (2/2)

The results

cipi


20

6. Conclusions

We successfully migrated the Java-based MEP to Scala

The evaluation tests show how the Scala-based MEP outperforms the Java-based MEP

Future work: Additional performance tests Focus on the Service Proxy (i.e., adpatation)

level

Final Match vs

cipi


21

Thank you for your attention!

Follow me on Twitter: @steccamiEmail: [email protected]

Michele Stecca

Scc2012 Scala

Technology

Transcript of Scc2012 Scala