1. Mobile Agent and Enterprise Architecture Integration Fulvio Di Marco Reti di Calcolatori LS aa. 2006/2007 Prof. Antonio Corradi Ref. Samuele Pasini

2. 2 Project Goals and Objectives □ The project aims to fill the lack of a software component integrated inside an Application Server infrastructure able to fully handle Mobile Agents platform logics. □ The adoption of mobile agents models inside enterprise software architectures and applications may further: ◊ the adoption of agents as proxies for a mobile user with capabilities of session and profile handling ◊ the ability by the side of the user to access common AS services by means of relocatable components ◊ the creation of mobility domains wider than the single cluster set in which the AS resides and enterprise boundaries

3. 3 Project Goals and Objectives □ Acquisition of a base knowledge relative to ◊ Application Servers (AS) architectures and bundled components and technologies (JMX, JMS, JNDI) ◊ Server-side component based programming ◊ Mobile Agents Technologies (SOMA) □ Extension of an existing mobile agents platform as integrating part of the AS infrastructure □ Development of mobile agents capable of interacting, through the new developed platform, with the AS and all its available components and services □ Accomplishment of resource handling politics able to trigger agents migration (Migration Manager)

4. 4 Technological Scenario JBoss Application Server Architecture □ JMX based microkernel architecture ◊ modular and dynamic solutions for managing, configuring and monitoring devices, applications, and service-driven networks ◊ It provides a common spine that allows you to integrate modules, containers, and plugins □ Fully compatible and Sun certified JMS 1.1 implementation ◊ It enables distributed communication that is loosely coupled, reliable, and asynchronous □ JNDI-based naming service ◊ unified interface to multiple naming and directory services

5. 5 Technological Scenario The Mobile Agents Paradigm and the Secure and Open Mobile Agents (SOMA) Platform □ Mobile Agent: composition of computer software and data which is able to migrate from one environment to another autonomously and continue its execution on the destination computer. □ Which benefits do agents provide for creating distributed systems? ◊ network load reduction ◊ asynchronous and autonomous execution ◊ dynamic adaptation (fault tolerance) ◊ naturally heterogeneous (seamless system integration) ◊ information dissemination (push) ◊ distributed information retrieval Locality Abstractions (scalability) □ place: the agent execution environment □ default place: acts as a gateway for interdomain routing. □ domain: a group of places,

6. 6 Enterprise Mobile Agent Platform Enterprise Application Integration JMX: framework for software integration. It provides a common spine that allows the user to integrate modules, containers, and plug-ins. Enterprise Mobile Agent Platform □ Architecturally structured in JMX MBean Services (core building blocks of the JBoss AS) □ Fully wired inside the AS infrastructure □ Full backport compatibility for agents □ Appearing among the services offered by the AS □ JMX Notification Framework □ JBossMQ Invocation Layer Services (JMS) □ JBossNS (JNDI) □ JBoss EJB container □ QOS awareness

7. 7 A MOM as a very fast glue... □ Fast and affordable integration of existing legacy systems (EAI) □ Asynchronous, loosely coupled communication □ Time ad location independence of components □ Latency hiding □ Scalability □ Event-driven systems □ Simplicity □ Optimal quality of service ◊ Persistence, guaranteed delivery, transactions

8. 8 Enterprise Name Services □ Enterprise context integration Extension of SOMA native Name Services (PNS, DNS) taking advantage of enterprise communication API provided by the AS □ These architectural choices ◊ Does not alter any locality abstraction hypothesis ◊ Are a fundamental preliminary remark to a further treatment of large scale distribution and QoS issues (fault tolerance, scalability, load balancing, persistence) □ Original SOMA platform PNS & DNS intrinsically suffer from the presence of single point of failures localized at default places ◊ Opportunity to benefit from the integration with the AS ◊ Distributing the connections across multiple servers in a cluster ◊ Storing messages in a relational database

9. 9 Enterprise Name Services JBoss Application Server JMS Provider JMS Provider Dom ainA Defa ult Plac e Root DNS Dom ainB Defa ult Plac e Plac e1 @ Dom ainA PN S Eve nts To pic Publisher on PNSEventsTopic Plac e2 @ Dom ainA Subscriber on PNSEventsTopic Subscriber on PNSEventsTopic 1. register 2. RegistrationResult 3. pns register msg 4. pns register msg Enterprise Place Name Service Enterprise Domain Name Service □ Star topology intra-domain and inter-domains mutual acquaintances □ Asynchronous “catch&return” semantics □ Point-to-Point & Publish/Subscribe models JBoss Application Server to subscribers: DomainB DomainC JBoss Application Server JMS Provider JMS Provider Dom ainA Defa ult Plac e Root DNS Dom ainB Defa ult Plac e Plac e1 @ Dom ainA Publisher on DNSEventsTopic Plac e2 @ Dom ainA 3. dns refresh msg 2. dns refresh msg 1. dns register msg Dom ainC Defa ult Plac e Subscriber for DNSEventsTopic Subscriber for DNSEventsTopic DN S Ro ot Qu eu e DN S Ev ent s To pic

10. 10 Inter-Domain Communications □ Location transparent message transfer mechanism □ Centralized JMS Provider ◊ Domain Public Remote Queues □ Local Domain Forward Queue □ Each Default Place implements a couple of Forwarding Units (Incoming & Outgoing) taking care of inter-domain communications □ Dynamically Resizable (via JMX) Pool of JMS Session Bridges □ Session Bridges make use of Enterprise PNS and DNS to locate destination queues JBoss Application Server JMS Provider JMS Provider Dom ainA Defa ult Plac e Dom ainB Defa ult Plac e Plac e1 @ Dom ainA PN S Eve nts To pic Plac e2 @ Dom ainB JMS Provider Public Remote Queues (inter-domain msgs) Domai nA Public Queue Domai nB Public Queue Do ma in For wa rd Qu eu e Forward Unit Forward Unit Pla ce2 Qu eu e forwardTo:Place2@DomainB

11. 11 Observer Agents Intra-Domain & Inter-Domains □ Collect and distribute informations about host resources (CPU charge, memory consumption, network load) to all the places belonging to the mobility domain distributed soft state □ destination place: selected using a simple and fair circular queue politic □ Observer Agents ◊ keep a personal and versioned dictionary of reports from visited places ◊ by querying appropriate API available at places ◊ obtain updated local reports, together with a version number managed by a counter local to the place ◊ trigger for a two-way synchronization between their personal dictionary and the one held by the hosting place

12. 12 Concluding Remarks □ Although limited to a limited number of physical machines in a LAN ($$$), performed tests have shown positive results, both logically and from a performance perspective □ Furthermore ◊ the robust scalability hypothesis which SOMA originally benefits from ◊ and scalability and reliability properties provided by AS related technologies □ are an encouraging starting point for testing efforts on large scale networks

13. 13 Future Developments The EAI process engaged by the works taking part to this project put the basis for: □ The implementation of additional EJB3/CORBA-based application services available to agents □ The substitution of mock/test-oriented resource sensors with appropriate and portable real sensors □ The extension of SOMA native security mechanisms in order to benefit from the security frameworks available in the enterprise context (JAAS) □ In-depth examination of basic themes concerning large scale distribution (fault-tolerance, load balancing, clustering, replication) □ The adoption of agents as proxies for a mobile user with capabilities of session and profile handling □ The ability by the side of the user to access common AS services by means of relocatable components