Model driven design and implementation of activity-based applications in Hermes

Abstract

— Hermes is an agent-based middleware structured as a component-based and 3-layered software architecture. Hermes provides an integrated, exible programming environment for design and execution of activity-based applications in distributed environments. By using work ow technology, it supports even a non expert user programmer in the model driven design and implementation of a domain speci c application. In this paper, after a description of Hermes software architecture, we provide a simple demo in biological domain and we show some real case studies in which Hermes has been validated.

-

B2

B1 C1 Pool of Workflow Executors (WEs) ilre Agent-Level Workflow WE -A-WE -B-WE -C- opm WE -A- WE -B- WE -CC F W

I. INTRODUCTION

Hermes [ 9 ] is an agent-based middleware, for design and execution of activity-based applications in distributed environments. It supports mobile computation as an application implementation strategy. While middleware for mobile computing has typically been developed to support physical and logical mobility, Hermes provides an integrated environment where application domain experts can focus on designing activity work ow and ignore the topological structure of the distributed environment. Generating mobile agents from a work ow speci cation is the responsibility of a context-aware compiler. Agents can also developed directly by an expert user using directly the Application Programming Interface (API) provided by Hermes middleware. The Hermes middleware layer, compilers, libraries, services and other developed tools together result in a very general programming environment, which has been validated in two quite disparate application domains, one in industrial control [ 6 ] and the other in bioinformatics [ 13 ]. In the industrial control domain, embedded systems with scarce computational resources control product lines. Mobile agents are used to trace products and support self-healing. In the bionformatics domain, mobile agents are used to support data collection and service discovery, and to simulate biological system through autonomous components interactions. This paper is organized as follows. Section II describes the Hermes Software Architecture. Section III provides a simple demo in biological domain. In Section IV, we present several projects in which Hermes middleware has been adopted. We conclude in Section V.

II. HERMES SOFTWARE ARCHITECTURE Hermes is structured as a component-based, agentoriented system with a 3-layer software architecture shown in Figure 1: user layer, system layer and run-time

Legend

Activity Agent

Tool Component

Start Stop Transition fork join

UserAgents WE -A- Response

Error SoapL

SA AIXO SA AIXO ServiceAgents (SAs)

Request

SA WE -A

WE -B

Email

SA Java Mail

API SOAPLAB

WSIF

SAXON layer. At the user layer, it allows designers to specify their application as a work ow of activities using the graphical notation. At the system layer, it provides a context-aware compiler to generate a pool of user mobile agents from the work ow speci cation. At the run-time layer, it supports the activation of a set of specialized service agents, and it provides all necessary components to support agent mobility and communication. The main difference between the run-time layer and the system layer is how agents function in each. ServiceAgents in the run-time layer are localized to one platform to interface with the local execution environment. UserAgents in the system layer are work ow executors, created for a speci c goal that, in theory, can be reached in a nite time by interacting with other agents. Afterwards that agent dies. Furthermore, for security UserAgents can access a local resource only by interacting with ServiceAgent that is the “guard” of the resource. It follows a detailed description of the main components and functionalities of each layer.

A. User Layer

The user layer is based on work ow technology and provides to users a set of programs for interacting with the worklow management system. There are two main families of programs: programs for specifying, managing and reusing existing work ow speci cations, and programs enabling administration and direct interaction with the work ow management system. The work ow editor is the program that supports the work ows speci cation by composing activities in a graphical environment. Hermes provides two editors, one is a plugin of the stand-alone JaWE [ 10 ] editor and the other is WebWFlow, a webbased editor. Both editors enable the speci cation of work ows by using XML Process De nition Language (XPDL) [ 14 ] a standard provided by the WfMC [ 12 ]. Activities used in a work ow are con gured by specifying input parameters and their effects are recognizable as modi cation of state variables or modi cation on the environment's status. Work ow editors enable the composition of both primitive and complex activities. A primitive activity is an activity that can be directly executed. Users can specify primitive activity without knowing the real implementation. A complex activity is an activity that must be speci ed before it can be used; as Figure 2 shows the speci cation of a complex activity could be a work ow of complex and/or simple activities. By using complex activities the speci cation of work ows is simpli ed because they enhance both hierarchical specication and reuse: we can use an already existing complex

Fig. 3. Outline of work ow compilation process in Hermes activity without caring of its speci cation. Users can use complex activities and stored work ows to increase productivity when specifying new work ows. Moreover, large libraries of both domain speci c primitives and complex activities can be loaded to specialize the editor for a speci c application domain.

B. System Layer

System Layer, on the middle architecture, provides the needed environment to map a user-level work ow into a set of primitive activities. The execution of these latter is coordinated by suitable model, they implement the activities at user level and embed implementation details abstracted from the execution environment. These primitive activities are implemented by autonomous software entities UserAgent able to react to the environment changes where they are executed. A compiler generates a pool of user mobile agents from the work ow specication. Due to the lack of space, work ow compilation process shown in Figure 3 will not discussed here and we refer to [ 4 ] for further details.

C. Run-time Layer

Run-time Layer, at the bottom of the architecture, provides primitives and services essential for agent mobility and resources access. The kernel is the platform for mobile computing which provides primitives for discovery, mobility, communication, and security. As already described, the overall structure of the system is very complex, it supports abstract speci cations that are mapped into a complex distributed and coordinated

ows of activities over a large-scale distributed system. In order to master this complexity and to support the reusability of existing artefact during the development of a middleware system for a speci c application domain, we designed Hermes kernel following a component-based [ 7 ] approach. Figure 4 shows the main components placed in the 3-Layered Architecture of Hermes Mobile Computing Platform. It follows a detailed description of components belonged to each layer.

1) Core Layer: It is the lowest layer of the architecture and contains base functions of the system, such as the implementation of the inter-platform communication protocols and agent management functions. This layer is composed of four components: ID, SendReceive, Starter and Security. The ID component, implements general identity management functions by managing a repository containing information about locally generated agents. This repository is accessed whenever we want to know the current position of an agent. The ID component is also responsible for the creation of the identi ers to be associated to new agents. These identi ers contain information about birthplace, date and time of the agent's creation. Agent localization is simpli ed by information contained directly in the ID, such as birth place. In fact, the birth place of an agent hosts information about agent's current location. A second important feature of the Core is the SendReceive component. This component implements low level inter-platform communication by sending and receiving messages and agents. By using traceability services offered by the ID component, SendReceive can easily update or retrieve the exact position of a speci c user agent. The Starter component processes any request for agent creation. This particular component, in fact, take an inactive agent (just created or migrated), and checks it for the absence of malicious or manipulated code. These agents, before activation, are dynamically linked to all basic services of the platform. During execution the agent is isolated from the Core Layer by the BasicService layer. The Security component, as mentioned above, checks for the presence of malicious code or manipulations within agent code.

2) BasicService Layer: This layer has ve main components: Discovery, Mobility, Genesis, Communication and Security Politics. The Discovery component searches and detects service agents. When a user agents wants to communicate with a service, it will ask the Discovery for the right identi er to use as the messages's receiver. The service detection strategy can be implemented in several ways; for example by a xed taxonomy or by an UDDI [ 5 ], commonly used in WebServices application domain. The mobility component enables the movement of code across platforms [ 11 ], it implements the interface used by the Agent component and it accesses to components of the Core layer to send, receive and load agents. It is important to note that real communication between different locations can be achieved only through Core's SendReceive component, and then migration is independent of the type of used transport. Mobility consists on copy the agent i.e. its code and its current state and send it to the destination platform where it will re-started in a speci c point (weak mobility). The local agent is destroyed. The Communication component makes possible to send and receive agent-directed messages both in an intra- and inter-platform context. Intra-platform messages are messages sent between agents and services residing in the same platform. Inter-platform messages are messages sent to agents residing in different platforms (our system does not allow for remote communication between user agents and service agents). The agent requesting the dispatch of a message does not need to know, effectively, where the target agent is; in fact, the ID is suf cient to post correctly a message. The Communication component uses one of the Security Policy's interfaces to ascertain whether the speci c UserAgent or ServiceAgent has the right privileges for communication. If an Agent is not authorized to use a service, the message is destroyed. Before accessing resources and services, an agent must authenticate itself. The identi cation is performed by sending a login message to a speci c ServiceAgent, as consequence the SecurityPolitics component jointly with the Communication component intercept the message and unlock the communication. The SecurityPolitics component centralizes control of permissions, protects services and resources from the user agents, and provides the administrator with an easy way to manage all permissions.

The last component of the service layer is the Genesis component that enables agent creation. A special case of agent creation is cloning that is performed when it is necessary to create a copy of an existing agent. The two copies differ only for the agent identi er.

3) Agent Layer: The Agent Layer is the upper layer of the mobile platform, the Agent Layer, contains all service and user agents. This component has not any interface, but it has only several dependencies upon the BasicService Layer. The Agent component provides a general abstract Agent class. UserAgent and UserAgent classes extend this abstract class. ServiceAgent consists of agents enabling access to local resources such data and tools. User agents execute complex tasks and implement part of the logic of the application. Java programmers can also develop UserAgents by using the API provided by Hermes Mobile Computing Library. Listing 1 shows a simple demo. A MkDuckAgent called “Della Duck” creates three sons Qui, Quo and Qua -lines 24 to 40- by cloning itself. After clonation each new agent start its behaviour calling “afterCloning” as initial method.

System . o u t . p r i n t l n ( ” H e l l o World ! ! ” ) ;

System . o u t . p r i n t l n ( ” I 'm D e l l a Duck ! ! ! ” ) ; p u b l i c void i n i t ( ) f r e c e p t i o n ( ) ; / / I e n a b l e t h e r e c e p t i o n

/ / o f messages f o r t h e f a t h e r By using “move” method -line 79- Qui, Quo amd Qua migrate to a Place different from where they were born. When they arrive in the new Place each one call the “afterMoving” -line 85- method. Then they notify to their mom their moving by using “sendMessageToUserAgent” -line 97- and “getMessageSynch” -line 101- methods. Figure 5 shows the nal results.

D. Software requirements

One of the main features of Hermes middleware is its scalability. The present version, HermesV2, is a pure Java application whose kernel requires about 120KB of memory and interoperates across a systems ranging from microprocessors to very power workstations. The Hermes Mobile Computing Platform is available under LGPL on Sourgeforge 1 Web Site.

III. MODEL DRIVEN DESIGN AND IMPLEMENTATION

OF ACTIVITY-BASED APPLICATIONS: A DEMO In the present post-genomic era, biological information sources are crammed with information gathered 1http://sourceforge.net/projects/hermes-project from results of experiments performed in laboratories around the world, i.e., sequence alignments, hybridization data analysis or proteins interrelations. The amount of available information is constantly increasing, its wide distribution and the heterogeneity of the sources make dif cult for bioscientists to manually collect and integrate information. In this section we present a demo of Hermes in the biological domain. In our example we want to nd similar DNA sequences to a given one in several databases using Basic Local Alignment Search Tool 2 (BLAST). In particular, in this demo we want to compare, using BLAST, the nucleotide sequence in FASTA format of a given entry identi cator with the sequences contained in the following databases:

Protein Data Bank (PDB) 3 SWISS-PROT 4

DDBJ 5

2http://www.ncbi.nlm.nih.gov/BLAST/ 3http://www.rcsb.org/pdb/ 4http://www.ebi.ac.uk/swissprot/ 5http://www.ddbj.nig.ac.jp/

The access to these databases is guaranteed by a set of Web Services. By using work ow editors a bioscientist can specify the logic order, as Figure 6 shows, of a set of domain-speci c activities without knowing the related implementation details. Each rectangle is an activity and each swimlane represents a UserAgent. As Figure 7-b shows, user can exploit a set of previous de ned domainspeci c activities by importing the proper library.

BlastnDDBJ Agent, BlastXSWISS Agent and BlastX

PDB Agent receive the nucleotide sequence from the BlastDemo Agent and throught an interation with WSIF ServiceAgent, they compare the received sequence with sequences in each database using BLAST. If no exceptions occur, BlastDemo Agent join partial results and send the nal document to user by email throught an interaction with the Email ServiceAgent. After saving this speci cation, you can reload -Figure 7-a-, compile -Figure 7-d- and execute -Figure 7-c and 7-e - the work ow previous de ned.

IV. SOME CASE STUDIES The Hermes middleware has been validated in several projects. It follows a brief case study description of Hermes application in some of them.

A. SI.CO.M project

In the SI.CO.M 6 project we have developed a prototype based on Hermes middleware for the traceability of ichthyic products. Generally the product is traced throught the updating of databases distributed along the main sites of the weaving factory. This approach is not ef cient because trace a faulty batch of products requires to query all databases, usually with an heterogeneous schema, of all sites interested in the production process. The proposed 6http://sicom.cs.unicam.it/ solution with the prototype named “TraceFish”, exploiting the agent-based technology, allows to move automatically the information about a single batch from a site to another of the weaving factory overcoming the limits of the classical client/server approach.

B. O2I Project

The Oncology over Internet (O2I) 7 project is aimed to develop a framework to support searching, retrieving and ltering information from Internet for oncology research and clinics. Hermes in the context of O2I project is called Bioagent 8, it supports the the design and execution of user work ows involving access to literature, mutation and cell lines databases.

C. LITBIO Project

The main objective of the Laboratory of Interdisciplinary Technologies in Bioinformatics (LITBIO) 9 is to create infrastructure capable of supporting challenging international research and to develop new bioinformatics analysis strategies apply to biomedical and biotechnological data. To satisfy the most bioinformaticians needs we have proposed a multilayer architecture [ 2 ] based on Hermes middleware. At the user layer, it is intended to support in-silico experiments, resource discovery and biological systems simulation. The pivot of the architecture is a component called Resourceome [ 8 ], which keeps an alive index of resources in the bioinformatics domain using a speci c ontology of resource information. A Work ow Management System, called BioWMS [ 3 ], provides a web-based interface to de ne in-silico experiments as work ows of complex and primitives activities. High level concepts concerning activities and data could be indexed in the Resourceome, that also dynamically supports work ow enactment, providing the related resources available at runtime. ORION [ 1 ], a multiagent system, is a proposed framework for modelling and engineering complex systems. The agent-oriented approach allows to describe the behavior of the individual components and the rules governing their interactions. The agents also provide, as middleware, the necessary exibility to support data and distributed applications. A GRID infrastructure allows a transparent access to the high performance computing resources required, for example in the biological systems simulation.

ACKNOWLEDGMENT

This work is supported by the Investment Funds for Basic Research (MIUR-FIRB) project Laboratory of Interdisciplinary Technologies in Bioinformatics (LITBIO).

V. CONCLUSION

As the demo presented shows, Hermes middleware provides an integrated, exible programming environment, whose user can easily con gure for its application domain. Hermes is structured as a component-based, agentoriented, 3-layered software architecture. It can con gured for speci c application domains by adding domainspeci c component libraries. The user can specify, modify and execute his work ow in a very simple way. Work ow is speci ed abstractly in a graphical notation and mapped to a set of autonomous computational units (UserAgents) interacting through a communication medium. The mapping is achieved by compiler that is aware not only of contents of a library of implemented user activities but also the software and hardware environment to executing them. By using work ow as suitable technology to hide distribution and on mobile agents as exible implementation strategy of work ow in a distributed environment, Hermes allows even to a not expert programmer a model driven design and implementation of a domain speci c activity-based application.

References

[1] M. Angeletti , A. Baldoncini , N. Cannata , F. Corradini , R. Culmone , C. Forcato , M. Mattioni , E. Merelli, and R. Piergallini . Orion: A spatial multi agent system framework for computational cellular dynamics of metabolic pathways . In Proceedings of Bioinformatics ITalian Society (BITS) Meeting , Bolgna, Italy, 2006 .

[2] E. Bartocci , D. Cacciagrano , N. Cannata , F. Corradini , E. Merelli, and L. Milanesi . A GRID-based multilayer architecture for bioinformatics . In Proceedings of NETTAB'06 Network Tools and Applications in Biology, Santa Margherita di Pula , Cagliari, Italy, 2006 .

[3] E. Bartocci , F. Corradini , and E. Merelli. BioWMS: A web based work ow management system for bioinformatics . In Proceedings of Bioinformatics ITalian Society (BITS) Meeting , Bologna, Italy, 2006 .

[4] E. Bartocci , F. Corradini , and E. Merelli . Building a multiagent system from a user work ow speci cation . In Proceedings of Workshop From Objects to Agents - WOA, 2006 .

[5] T. Bellwood , L. Cle´ment, D. Ehnebuske , A. Hately , M. Hondo , Y. L. Husband , K. Januszewski , S. Lee , B. McKee , J. Munter , and C. von Riegen. UDDI version 3 .0. Published speci cation, Oasis , 2002 .

[6] D. Bonura , F. Corradini , E. Merelli, and G. Romiti. Farmas: a MAS for extended quality work ow . In 2nd IEEE International Workshop on Theory and Practice of Open Computational Systems . IEEE Computer Society Press, 2004 .

[7] D. Bonura , L. Mariani , and E. Merelli . Designing modular agent systems . In Proceedings of NET . Object DAYS , Erfurt, pages 245 – 263 , September 2003 .

[8] N. Cannata , E. Merelli, and R. B. Altman . Time to organize the bioinformatics resourceome . PLoS Comput Biol ., 1 ( 7 ):e76, 2005 .

[9] F. Corradini and E. Merelli . Hermes: agent-based middleware for mobile computing . In Mobile Computing , volume 3465 , pages 234 – 270 . LNCS , 2005 .

[10] Enhydra . Jawe. http://jawe.enhydra.org/, 2003 .

[11] A. Fuggetta , G. Picco, and G. Vigna. Understanding code mobility . IEEE Transaction of Software Engineering , 24 ( 5 ): 352 – 361 , May 1998 .

[12] D. Hollingsworth . The Work ow Reference Model , January 1995 .

[13] E. Merelli , R. Culmone , and L. Mariani . Bioagent: a mobile agent system for bioscientists . In NETTAB Workshop on Agents Nd Bioinformatics , Bologna, July 2002 .

[14] WfMC . Xml process de nition language (xpdl) . WfMC standard, W3C , October 2005 .