Hypersonic { Model Analysis as a Service

Abstract

Hypersonic is a Cloud-based tool that proposes a new approach to the deployment of model analysis facilities. It is implemented as a RESTful Web service API o ering analysis features such as model clone detection. This approach allows the migration of resource intensive analysis algorithms from monolithic desktop modeling tools to a wide range of mobile and Web-based clients. As a technology demonstrator, a Web application acting as a client for the Hypersonic API has been implemented and made publicly available.

Introduction

The vast majority of modeling tools are currently deployed as rich client desktop applications. Some tools, such as those based on the Eclipse Modeling Framework (EMF), bene t from rich plug-in ecosystems. Others, such as the MagicDraw modeling environment [ 8 ], o er built-in access to remote resources (e. g. a version control server). However, recent years have seen the rich client architecture demonstrating its limitations, causing many areas of computing to consider more exible alternatives. A notable example are Cloud-based architectures, which involve the deployment of computationally intensive tasks to a centralized and fully transparent shared pool of con gurable computing resources. Lightweight clients access these resources via protocols such as RESTful Web services [ 4 ], giving rise to the \Software as a Service (SaaS)" paradigm. The adoption of this paradigm in the area of modeling tools is so far remarkably limited.

In this context, we have proposed Hypersonic [ 1 ], a RESTful Web service API o ering model analysis facilities including clone detection, model di erence computation, and model size computation. Model analysis is a particularly suitable application area for a service-oriented architecture, since analysis algorithms often demand extensive resources. The service-based approach to modeling tools is illustrated in Fig. 1: a subset of the model processing features deployed locally in a rich client architecture may instead be o ered as services by several providers, while existing remote components such as model repositories remain una ected.

This approach bene ts all the involved stakeholders. For modelers, it supports access to scalable computational resources running the latest version of the modeling tool, as well as the option to mix-and-match services o ered by di erent providers. For tool providers, it facilitates cross-device support and mitigates the distribution of unlicensed software. Furthermore, entry barriers for new providers are lowered. For IT administrators, a service-based architecture brings a reduction in installation and maintenance workloads. Finally, the Model-Based Software Development (MBSD) community can expect improved standards compliance and an accelerated knowledge transfer between research and industry.

The comparatively few drawbacks of a service-based approach are caused by its distributed nature. Uploading large models may create a performance bottleneck, while new security and privacy aspects come into play. Nevertheless, these drawbacks are common to most Software as a Service (SaaS) solutions, and have not undermined this architecture style's acceptance.

Client

Provider 1

Client

Provider 1

. . .

Client

Provider k

. . .

Provider x

Provider y

Provider 1

Provider 1

d n e g e L

Component suitable for deployment to remote server

Component suitable for deployment to local client

Hypersonic

The Hypersonic API The architecture of the Hypersonic API is presented in Fig. 2. Following the REST architectural style, it exposes resources for clients to interact with via HTTP requests. The API currently supports MagicDraw UML models (support for EMF and Simulink [ 12 ] models is also planned). API response messages contain the analysis results encoded as JSON [ 5 ] documents.

The Hypersonic API is implemented as a wrapper around the MACH model analysis engine [ 11 ], though from a conceptual standpoint any other model analysis engine may be used. The RESTful API component handles HTTP interactions with clients, delegating processing tasks to the MACH engine. All uploaded API client .xmi HTTP HTTP .json

XMI2PL

.xmi

RESTful

API call(…) […] .pl

MACH

.pl

Model warehouse

models pass though the XMI2PL component, which performs a format translation from the MDXML format to the Prolog-based le format described in [ 10 ]. Once translated, models are stored in a dedicated warehouse. The MACH component supports operations such as clone detection, model di erencing, model size computation, and basic model querying [ 11 ]. These operations can be applied to models stored in the warehouse. Analysis results are encoded by the RESTful API component as JSON documents to be consumed by the API client.

All processing components are executed inside a single instance of the SWIProlog [ 13 ] runtime. The compact representation of models as fact databases facilitates the e ective in-memory processing of large models. Since some analysis operations (e. g. model size computation) are embarrassingly parallel, splitting very large models into several fact databases and executing several SWI-Prolog instances in parallel is possible, though currently not supported. As a technology demonstrator, a simple, mobile device friendly Web application acting as a client for the Hypersonic API1 has been implemented. The application, written entirely in client-side JavaScript, currently supports uploading a local model to the Hypersonic model warehouse and requesting a clone detection report, which it subsequently displays in tabular form. The sample API client exempli es our vision for Web service driven modeling tools: using Web 2.0 technologies and standards (REST, JavaScript, JSON) to enable advanced model analysis operations to be performed outside the constraints of the desktop and of traditional modeling environments. 1 The application is available at http://www.compute.dtu.dk/~rvac/hypersonic. Due to the increase in size of industrially relevant models, as well as the increase in complexity of the operations performed on them, the need for service-based, Cloud-enabled modeling tools has become apparent [ 7 ]. The main driver in this direction is the promise of performance and scalability gains for modeling activities such as model warehousing [ 9 ], querying [ 6 ], and transformation [ 3 ].

That being said, service-based model analysis has yet to receive signi cant attention in the literature. The closest related proposal is the EMF-REST project [ 2 ], aimed at automatically generating RESTful Web service interfaces for EMF models, much like existing EMF tools generate Java APIs for such models. However, while it does provide basic model manipulation operations, EMF-REST is not primarily designed as a model analysis tool. 4

Ongoing Developments

The Hypersonic API is undergoing development with the aim of reaching feature parity with the MACH command-line model analysis tool. Once this is achieved, the API will be deployed to a private Cloud platform. We plan to eventually extend the API operations beyond model analysis, into the realm of model querying and model transformation. In parallel, we will work towards creating an ecosystem of API clients, including smartphone/tablet apps and modeling tool plug-ins (see Fig. 4). For these proposals to become practical solutions, security aspects such as user authentication and model access control must be addressed.

Smartphone app

Web app Desktop client Modeling tool plug-in HTTP

Hypersonic API

MACH

References

1. Vlad Acretoaie and Harald Storrle. Hypersonic: Model Analysis and Checking in the Cloud . In Proc. 2nd Ws. Scalability in Model Driven Engineering (BigMDE'14) , volume 1206 of CEUR-WS , pages 6 { 13 , 2014 .

2. Jordi Cabot . EMF-REST . http://emf-rest.com, retrieved 26 .08. 2014 .

3. Caue Clasen , Marcos Didonet Del Fabro , and Massimo Tisi . Transforming Very Large Models in the Cloud: a Research Roadmap . In Proc. First Intl. Ws. ModelDriven Engineering on and for the Cloud (CloudMDE'12) , pages 3 { 12 , 2012 .

4. Roy Thomas Fielding. Architectural Styles and the Design of Network-based Software Architectures . PhD thesis , University of California, Irvine, 2000 .

5. Internet Engineering Task Force (IETF) . IETF RFC 7159 : The JavaScript Object Notation (JSON) Data Interchange Format , 2014 .

6. Benedek Izso , Gabor Szarnyas, Istvan Rath, and Daniel Varro. IncQuery-D: Incremental Graph Search in the Cloud . In Proc. 1st Ws. Scalability in Model Driven Engineering (BigMDE'13) , pages 4 :1 { 4 : 4 , New York, NY, USA, 2013 . ACM.

7. Dimitrios S. Kolovos , Louis M. Rose , Nicholas Matragkas, Richard F. Paige, Esther Guerra, Jesus Sanchez Cuadrado, Juan De Lara, Istvan Rath, Daniel Varro, Massimo Tisi, and Jordi Cabot . A Research Roadmap Towards Achieving Scalability in Model Driven Engineering . In Proc. 1st Ws. Scalability in Model Driven Engineering (BigMDE'13) , pages 2 :1 {2: 10 , New York, NY, USA, 2013 . ACM.

8. NoMagic, Inc. MagicDraw UML 17.0 .3. http://www.nomagic.com/products/ magicdraw, retrieved 26 .08. 2014 .

9. Javier Espinazo Pagan , Jesuss Sanchez Cuadrado, and Jesus Garc a Molina. Morsa: A Scalable Approach for Persisting and Accessing Large Models . In Proc. 14th Intl. Conf. Model Driven Engineering Languages and Systems (MODELS'11) , volume 6981 of LNCS , pages 77 { 92 . Springer Berlin Heidelberg, 2011 .

10. Harald Storrle. Towards Clone Detection in UML Domain Models . J. Softw. Syst. Model. , 12 ( 2 ), 2013 .

11. Harald Storrle. UML Model Analysis and Checking with MACH . In Proc. 4th Intl. Ws. Academic Softw. Development Tools and Techniques (WASDETT'13) , 2013 .

12. The MathWorks , Inc. Simulink. http://www.mathworks.se/products/simulink/, retrieved 26.08 . 2014 .

13. Jan Wielemaker , Tom Schrijvers, Markus Triska , and Torbjorn Lager. SWI-Prolog. Theory and Practice of Logic Programming , 12 ( 1-2 ): 67 { 96 , 2012 .