Collaborative Engineering Approach Towards IP-Based SoC Design
Adam Pawlak & Piotr Penkala, Silesian University of TechnologyHåvard D. Jørgensen, Active Knowledge ModelingPawe³ Fra, Silesian University of TechnologyWojciech Sakowski, Evatronix SA
The paper introduces a visual knowledge-based modelling platform and a system enabling integration of distributed design tools that have been deployed in a distributed collaborative scenario of IP components design. The approach integrates concepts of active knowledge models well suitable for modelling business processes, with engineering workflows that can represent IP-based SoC design flows. Design workflows are realised in the system based on Tool Registration and Management Services (TRMS) that constitutes a core of the engineering collaborative infrastructure that enables distributed design. The system assures information security (including user authorization, data and transfer encrypting), remote administration of users and tools, and some support for distributed inter-organization workflows. The collaborative infrastructure has been applied to IP component design that required collaboration among two dispersed SMEs. The paper presents relevant results of the EU project MAPPER.
1. CHALLENGES IN COLLABORATIVE DESIGN OF ELECTRONIC SYSTEMS
Todays design and verification processes are inherently complex, distributed and interdisciplinary. Furthermore, they are characterized by use of heterogeneous tools and extensive reuse of pre-designed configurable IP components. Rising complexity and heterogeneity of systems is forcing companies to establish more and more mutual collaboration links. This tendency for rising number and range of collaborative links is globally visible. The emerging new paradigm of distributed work is referred as collaborative engineering  which is understood as an innovative method for product design and development that integrates widely distributed engineers for virtual collaboration. Collaborative engineering becomes feasible due to recent advancements in ICT. It is however often restricted to engineering groups from large global companies, but even there support for collaboration between dispersed companys sites remains limited. One of the reasons for these limitations is that engineering knowledge sharing in distributed collaborative design processes usually takes place in established partnerships only, where mutual trust exists and appropriate agreements have been set up.
Sometimes, especially in case of SMEs, collaboration links take a more rigid form of collaborative networks (often called virtual organisations) . When applied to distributed engineering processes they are referred to as Global Engineering Networks (GEN)  or collaborative engineering networks . Management of tools in such networks requires new approaches that enable a designer to create virtual design environments that integrate tools and services (e.g., compilers, simulators, logic and behavioural synthesis tools, physical design tools, test equipment, product data management systems and/or databases) that are distributed over different collaborative network sites and platforms .
Internet-based collaborative environments for applications in Electronic Design Automation (EDA) have been a target of academic R&D since a decade. The examples of the most relevant academic solutions are: WELD , JavaCAD , REUBEN , ASTAIR  and MOSCITO . It should be noticed that, despite many research efforts, including those mentioned above, widely accepted Internet-based tool integration technology is still not available. Furthermore, the existing distributed collaborative engineering infrastructures do not sufficiently support such requirements, like intellectual property protection, information security (including user authorization, data and transfer encrypting), communication through firewalls, support for distributed inter-organization workflows, remote administration of users and tools, and as already mentioned, engineering knowledge sharing.
Further challenges related with inter-company collaboration, and especially collaboration in networks formed by SMEs are related to:
inter-company secure collaborative infrastructures enabling dynamic partnership in the network;
network-aware design methodologies including methodologies for participatory design that will enable smooth involvement of users in design processes;
adequate eBusiness models enabling access to engineering services and functionality of required eBusiness transaction over the network.
With the MAPPER EU project  and the former E-Colleg project  approach which has been applied to electronic system design, and more specifically to IP-based System-on-a-Chip (SoC) design we address some of the above challenges. MAPPER solutions relevant for applications in electronic systems design, based on the Active Knowledge Modelling Platform  and the Tool Registration and Management Services (TRMS) environment  have been introduced in the paper.
2. THE POTENTIAL OF ACTIVE KNOWLEDGE MODELLING IN ENGINEERING
Active knowledge models (AKMs)   are visual representations of unfolding and dynamic business knowledge. The models are used actively to customize and adapt the IT infrastructure, and the models are executed through process enactment and rule engines. Active knowledge models differ from conventional model-driven architectures (MDA) in that they primarily capture user knowledge about business realities, rather than technical information about how the computerized support systems work. AKM thus lets evolving business needs directly control the IT infrastructure. Its interactive model execution paradigm, which flexibly merges automatic reasoning and manual decision making, creates IT support for the innovative design processes at the core of the business competitive advantage, not just for the administrative support processes.
The analysis of potential for deploying AKM concepts in engineering of electronic systems reveals the following observations:
Active Knowledge Model (AKM) of a design process can serve as an active and systematically updateable knowledge base on the design environment with: relevant design processes including those used for quality assurance, required human resources (engineering expertise profile), and tools. This can simplify transfer of design knowledge around a particular design phase or design task to new employees;
AKM defined as a design workflow, when executed can support (semi-) automation of design processes;
AKMs in a very natural way can model electronic products at various levels of abstraction, from simple Intellectual Property components to complex heterogeneous systems. Here, configurability of active knowledge models supports design of families of products. Being deployed in teams of engineers, they may be composed to form models of even more complex design processes.
2.1. Representation and Execution of Design Processes as AKMs
Models are generally defined as explicit representations of some portions of reality as perceived by some actor. Active models, on the other hand, are also actively used during the operation of the system. What does it mean that a model is active? First of all, the representation must be available to the users of the information system at runtime. Second, the model must influence the behaviour of the underlying computerized system. Third, the model must be dynamic and reconfigurable, users must be supported in changing the model to fit their local reality, enabling tailoring of the system's behaviour.
Figure 1 presents an example of an IP component development process represented as a visual active knowledge model. The active model comprises a set of hierarchical visual objects that represent the processes of specification, development and productization (product preparation), the definition of information technology (IT) infrastructure, company organisation charts, design data and engineering knowledge structure. All development processes form a common design automation workflow. Each block, describing the main development stages, contains several sub-blocks that bring detailed description of engineering tasks to be performed. Such a workflow exhaustively defines the sequence of tasks and relations between objects. For each task there are the input and output parameters defined.
The other elements of the model represent not processes but the organisational aspects of the IP component development. The model of IT infrastructure gives the information on what tools, that support design automation, office work and automation, are available for use. It also presents existing hardware resources, e.g., workstations, test equipment etc. The model of organisations involved in IP component development describes the structure of companies and defines the roles necessary for performing the tasks. The roles are linked to the actually available personnel. The model also gives the information about geographical locations of collaborating organisations. Another important objects of the model describe the elements and structure of design data and represent additional engineering knowledge content, e.g., a set of best engineering practices. Each visual object in the model can be related to the other one enabling creation of a valuable knowledge representation. For example, a particular development task can be linked to the human and not human resources necessary to perform it.
Click here to view a larger, more detailed version
Figure 1. Example of AKM deployed to IP component development.
2.2. Design Workflows in TRMS
A general architecture of TRMS includes three components: Global Tool Lookup Service (GTLS), Tool Servers (TSs), and Client Applications. The main GTLS component is responsible for registration and modification of data on users and their privileges, elements of the system, as well as information on accessible tools and machines that make them available. GTLS also manages security policy of the whole system.
TRMS enables secure data transfer with authentication and authorization of users, as well as it includes security management mechanisms that allow an administrator to monitor users' activity and to execute a proper security policy. TRMS uses the HTTP SSL channel (by default) for communication between components. In addition, all sensitive data can be encrypted and digitally signed by a sender for an improved security level. Using standard Internet protocol is useful for a networks protected by firewalls.
Figure 2. Distributed tools invocation with TRMS.
Figure 2 shows a typical scenario of TRMS use in a design automation domain. A user executes a workflow comprising a sequence of required design tasks. The TRMS Client interface helps to define the workflow and control execution of tasks in a graphical way. In the presented example, the workflow consists of a sequence of three design tasks. Two of them are executed on Tool Server 1 (e.g., compilation and simulation) and the third one on Tool Server 2 (e.g., synthesis). The appropriate tool servers, corresponding to the required design services, have been picked up with a help of the GTLS service. A task, being a part of the workflow available at the GUI level, is a representation of a service that is executed at the Tool Server.
2.3. Model-Configured Task Execution in the AKM Platform
In order to connect the high level business process models of the design flows to the concrete tasks and workflows that are made available through TRMS, the atomic tasks of the business processes must be connected to TRMS workflows or tasks. Then the AKM task execution engine will know which lower level tool to invoke for each process step. Interactive tasks are triggered using the parameterised URL interface of the TRMS applet, while automatic workflows may also be invoked behind the scenes through the TRMS web service interface. The AKM task execution engine supports both these integration mechanisms.
The data and parameters needed as input to TRMS, are extracted from the AKM models. By defining mapping relations between model elements and service input and output parameters, any kind of modelled content may be used by or produced by lower level services. Any kind of model information, whether concerning documents, organisations and people, processes and task, product structures etc., can be configured to be used by the services. The integration of TRMS and AKM is thus flexibly model-configured, and may be dynamically adapted to meet local needs, thus supporting creative design.
2.4. Integration of Task Execution and TRMS Workflows
Communication between components of the TRMS environment (Figure 2) is based on Web services. A simple set of services is made available by TRMS in order to support integration with other collaborative systems. This set of services has basic functionality and a simplified data structure. Due to that integration with other collaborative design environments is straightforward. TRMS Web services offer currently support for user authentication, authorisation as well as search for predefined tasks or groups of tasks (workflows). Tool registration is possible, with the use of the full TRMS application version only.
The search process enables discovery of an appropriate tool (or a group of integrated tools) for executing of a concrete task. Upon a task (or workflow) selection it is possible to invoke it through the use of the simplified TRMS client in a form of a parameterised applet. This applet enables invocation of a concrete task (or workflow) and control of its execution. In case of a workflow, a user has a possibility to invoke one of declared tools, or a workflow. A user can monitor the invoked task (or workflow) with the monitoring facilities offered by the TRMS client.
3. MAPPER APPROACH TO DESIGN OF ELECTRONIC SYSTEMS
The presented approach towards distributed collaborative IP component design integrates visual knowledge modelling (AKM) with the distributed tools management and invocation (TRMS).
3.1. Use of AKMs to Collaborative Design of IP Components
AKMs have been deployed to collaborative design of an IP component in the frame of the MAPPER project in an experiment that integrates two IP components development companies from Germany and Poland that aim at designing an advanced USB IP component. An IP component is required for hardware implementation of standard serial communication protocols. Obviously, this design needs to fulfil constraints on time to market, security of design data, and the customer-specific final product functionality. This IP design is being performed in a truly distributed collaborative engineering network that is being built upon the MAPPER infrastructure. Each node in this network constitutes an engineers workspace equipped potentially with design tools or just a remote tool that processes automatically design data.
The distributed design process commences with an initial, informal specification of the Hard IP design that is agreed upon by both companies and their customers. Once a precise specification exists both companies define their design workflows as active knowledge models, and define interfaces. These AKM-based workflows span over: specification, development, verification, and product preparatory phases. In the following, actors responsible for particular design phases in the design flow, technologies for components manufacturing, and tools to be used, are identified. The common design workflow defines all design steps at both companies that are needed for designing and production of the USB IP component. This common design workflow is a result of numerous consultations between managers and engineers of both companies that are performed using also the support of the MAPPER infrastructure.
3.2. Aligning Multiple Design Flows
The two companies first modelled their core design processes. The design flow represented as a visual model in Figure 1 comprises sets of design tasks performed in three geographical locations, i.e., one in Germany (described as A) and two others in Poland (company branches B and C). The design object is a mixed-mode component. Design processes are split among the team members, taking the engineering competences into account, in such way that an analogue design takes place in location A, a digital design in B, and a board-level testing in C. The common design flow starts with the design specification phase, goes through the digital and analogue design and verification phases, the design integration and implementation phase, the design testing and finalises in the design productization (product preparation) phase. The main design phases are depicted in the bottom part of the diagram (Figure 1).
The visual model developed for an IP component design comprises also a wide spectrum of information related to the current joint product, namely information on: the internal organisation of involved companies (e.g., company structure, locations, human resources, staff competence skills), the available IT infrastructure (e.g., design automation, administration, and office tools), the current project organisation (e.g., project responsibilities), the detailed structure of the joint product, and the project plan (e.g., management and design workflows). The important part of the model are relations that link model objects. The visual model comprises a significant amount of elements, i.e., objects and relations, not easy to grasp all at once, therefore it is necessary to focus only on the portion of information at the time. Within the AKM model browser it is possible to control visibility of model elements by creating different views with selected model components only, or controlling the visibility of particular model objects. An example of the model scoping is presented in Figure 1. The block P3.2/2 Design and Verification that takes the second position in a top level design flow, is zoomed in to show the internal hierarchical structure of design sub-tasks.
The visual model describes formally the design methodology, defining the sequence of tasks that have to be executed to reach the project target. It also constitutes a set of design guidelines for engineers and managers. The visual model gives the foundations for the development of TRMS workflows. The TRMS workflow enabled designers to easily access the design tools, namely, an HDL simulator and a synthesis tool, installed at the remote location. In the performed experiments, the test engineer in location C, validating the FPGA implementation of the IP component at a board level, could easily solve simple design bugs discovered in a digital controller part. He/she could correct a HDL code of the controller and, via TRMS, launch the sequence of EDA tools located in B, and generate the fixed FPGA implementation. In result, the process of board testing did not depend upon neither the availability of on-site design tools nor the active participation of additional engineers in location B. This iteration took less time than in the case without remote tool invocation. On the other hand additional design skills are required from the test engineer.
4. CONCLUSIONS AND FUTURE WORK
The paper introduces basic concepts of a visual knowledge modelling method called AKM Active Knowledge Modelling. It demonstrates how these concepts and the underlying collaborative framework can be applied to distributed design processes of a complex IP component. The collaborative framework encompasses the TRMS system that can enable invocation of distributed design tools across the Internet. The following conclusions point to some already experienced, as well as potential advantages of the presented approach:
As the AKM approach is engineering domain independent it can be applied to various modelling tasks that cover heterogeneous design processes often required in design of new complex systems;
Visual AKM representation of design processes and workflows is natural to designers. Still work needs to be invested in elaboration of more electronic engineer specific profiles and views at the AKM platform;
Visual AKM representation of structural properties of electronic products should also be well perceived by engineers. Visual representation of behavioural functionality of electronic products is less evident to engineers, who tend to express this functionality in a textual form of hardware description languages. This aspect of electronic component modelling using AKM approach requires further investigations;
There is a big potential for deploying the AKM technology for engineering task patterns representation. Task patterns can support automation of various engineering tasks, like virtual meetings, or engineering service search in collaborative networks, that are hardly handled by traditional EDA environments;
TRMS demonstrates its advantages for secure integration of distributed tools. Its workflow execution engine needs however to be improved in order to support invocations of complex sequences of design tasks.
Integration into one homogeneous model of various aspects of business processes, e.g. design processes, workflows, design environments with human and tool resources, is an important step towards knowledge-based engineering. It is believed that deployment of the MAPPER technology to distributed collaborative design of SoCs using IP components will result in an improved re-use, management and protection of IP components, as well as knowledge-enhanced design processes.
 ASTAI® Manual v. 2.2. C-Lab, Univ of Paderborn, Germany, 2000.
 Camarinha-Matos Luis M., Afsarmanesh H.: The emerging discipline of collaborative networks, Luis Camarinha-Matos (Ed.) "Virtual Enterprises and Collaborative Networks", Kluwer Academic Publishers, 2004.
 Chan, F.; Spiller, M.; Newton, A.R.: WELD an Environment for Web-based Electronic Design, Proceedings of the 35th DAC, 1998.
 Cutkosky M, Tenenbaum J, Glicksman J.: Madefast: an exercise in collaborative engineering over the Internet, Communications of the ACM. 1996, vol. 39, no. 9.
 Dalpasso M, Bogliolo A, Benini L.: Specification and validation of distributed IP-based designs with JavaCAD, Proc. DATE, 1999.
 E-Colleg, MAPER and TRMS web pages: http://www.ecolleg.org, http://mapper.eu.org, http://mapper.eu.org/miug/trms/
 Fra P, Kostienko T, Magiera J, Pawlak A, Penkala P, Stachañczyk D, Szlêzak M, Witczyñski M.: Collaborative infrastructure for distance - spanning concurrent engineering, as .
 Jørgensen, H.D.; Karlsen, D.; Lillehagen, F.: Collaborative Modelling and Metamodeling with the Enterprise Knowledge Architecture, Enterprise Modeling and Information Systems Architectures, An Int. Journal, Vol. 1, No. 1, German Informatics Society, 2005.
 Lavana, H.; Khetawat, A.; Brglez, F.; Kozminski, K.: Executable Workflows : A Paradigm for Collaborative Design on the Internet, Proceedings of the 34th DAC, 1997.
 Brglez F.: E-Design Concepts and Practice: An Overview, IP SoC Grenoble, 2001.
 Lillehagen, F.: The Foundation of the AKM Technology, in Jardim-Gonçalves, Cha & Steiger-Garção (eds), Concurrent Engineering, Enhanced Interoperable Systems: Rotterdam: Balkema, 2003.
 Radeke E.: GEN - Global Engineering Networking, Proc. Conf. on Integration in Manufacturing, 1998, Goteborg.
 Schattkowsky, T.; Mueller, W.; Pawlak, A.: Workflow Management Middleware for Secure Distance-Spanning Collaborative Engineering, In L. Fischer (ed.) The Workflow Handbook 2004, WfMC, Lighthouse Point, USA, 2004.
 Schneider A, Ivask E.: Internet-based collaborative system design using Moscito, Proc. Workshop on Challenges in Collaborative Engineering 15-16.04.2003, Poznañ, Poland.
 Siekierska K. et al.: Distributed collaborative design of IP components in the TRMS environment, Microelectronics Reliability, Elsevier Journal, vol. 46 (2006), 5-6.
 Witczyñski, M.; Pawlak, A.: Virtual Organisations Enabling Net-based Engineering, Challenges and Achievements in E-business and E-work, Stanford-Smith B. et al. (Eds.) IOS Press, Amsterdam, 2002.
Contact Evatronix S.A