COMPARATIVE STUDY OF JADE AND SPADE MULTI AGENT SYSTEM.

Geetha Radhakrishnan 1 , Chithambaram.V 2 , and Shunmuganathan K.L 2 . 1. Asst. Prof., Kerala Agricultural University, Kerala. 2. Principal, Dhanalakshmi College Of Engineering, Chennai. 3. Research Centre, Dhanalakshmi College of Engineering, Chennai. ...................................................................................................................... Manuscript Info Abstract ......................... ........................................................................ Manuscript History Received: 12 September 2018 Final Accepted: 14 October 2018 Published: November 2018

The rest of this paper is divided as follows. In Section 2 we begin our discussion for agent based cloud security by analysing relevant works done in the area of multi-agent systems, FIPA and SPADE. These studies introduce us to the purpose of multi-agents and how they have been used in various sectors. We also take a look at FIPA standards and innumerable efforts to improve these guidelines. The prevailing agent platforms along with their various drawbacks are also explained. Section 3 presents the architecture of SPADE and JADE platforms. It discusses the step by step development of agents using SPADE and JADE, and gives an outline of the execution of these recommended agents. Section 4 examines the observations and provides the system evaluation. Section 5discusses the inferences, implications and prospective improvement areas to augment this archetype.

ISSN: 2320-5407
Int. J. Adv. Res. 6(11), 1035-1042 1036 Background Multi-agent systems have found use in varying fields through their logical, technical, practical, efficient methods to decipher and resolve challenges [7] - [12]. Along with the ability to interact with each other and with their alternative actions [13], they also have self-healing protection systems [14]. This society of intelligent agent systems capable of adjusting to situational variations, interact and exchange knowledge to achieve a common goal without external interventions, is a Multi-Agent System (MAS) [10]. MAS may contain a combination of humans and agents. Agents in a MAS are autonomous and have only a local view of the system. Also, no agent in the MAS is assigned with governing or supervisory authority.
Implementations of agents are based on Foundation for Intelligent Physical Agents (FIPA) standards [15]. FIPA is the organization that establishes standards for computer software, interacting agents and agent-based systems. According to FIPA, an agent is a computational process that implements autonomous functionality of an application. Agents are autonomous, which means that they are independent, self-governing and have control over their own actions. They are reactive and respond to variations in the environment. They learn from previous experiences and make impromptu tactical adjustments and take initiatives to achieve their goals as they are goal-directed, thereby displaying their pro-activity and flexibility [16]. Agents are social and interact with both humans and other agents. They are also mobile and are able to move with ease from one machine to the other. Their robust nature ensures continuity and vigorous fault recovery to realize goal achievements [17].
The FIPA Agent Platform (AP) follows the FIPA reference model. This comprises of Agent Management System (AMS), Directory Facilitator (DF), Agent Communication Channel (ACC) and the Internal Platform Message Transport (IPMT).The DF provides an indexed almanac to the services provided by other agents registers and deregisters the agents. AMS also registers and deregisters agents, provides an alphabetically filed directory of agent descriptions, supervises and controls access to the AP and delivers life-cycle management facilities. The ACC provides inter-agent communication both internally and traversing through platforms. IPMT provides message forwarding services within a platform [18], [19]. The semantics/ontologies aid in the definition of names and data types to be used in messages and their meanings. Messages in the colloquy are explicitly interpreted by all involved participants [21]. FIPA standards has played a major role in the development of Multi Agent Systems MAS [22], [23].
Several agent development platforms have evolved and many more are being conceived. Selection of the ideal agent to suit our purposes is an important factor in agent development. Based on the surveys of agent platforms [24] - [27], we understand that most platforms are Java based. Of these very few have operating systems and are still active. It is clear that only 5 of them follow FIPA standards and an even lesser number has got good agent communication language.

Fig. 2:-Agent Platforms features using Java and Python
Of the agent platforms that were reviewed, Java Agent Development Environment (JADE) and Smart Python Agent Development Environment were found to be ideally suited for agent development. The agents developed using SPADE run on distributed machines and are capable of communicating with each other. Each agent when created is assigned with a Global Unique IDentifier (GUID) and a Transport Address (TA) so as to enable registration with the white paper services. APIs are provided with the task to create, suspend, resume, freeze, migrate, clone and kill agents. , which facilitates the white pages and default yellow pages of the system, supervises the container table, which is the registry of object references and transport addresses of all the container nodes and manages the Global Agent Descriptor Table  (GADT), a registry of all agents in the platform. Platform Connection: SPADE agent has in-built jabber technology and so, when an agent is connected to the SPADE platform, it gets automatically registered, and remains active, persistently communicating with the agent platform. Message Disseminator: This is the component that ensures delivery of a message to the correct recipient or agent. Agent behavior: This is the task executed by the agent repeatedly. An agent can have several behaviours and each behavior has a message template attached to it. This template determines the kind of message that a behavior should receive. There are some pre-defined behavior types postulated by SPADE and these aid in the implementation of the tasks assigned to an agent. Java Agent Development Environment is a robust and efficient agent platform, developed using Java language, and follows the FIPA standards. Due to its JAVA base, it can effortlessly network with Java implementations and has the ability to be distributed across machines running on different Operating Systems. JADE includes an active run-time environment where agents can live, a library of classes that programmers have to develop, a graphical tools suite for administering and monitoring the running agents [31].

JAVA
Active Jade environments are called containers. Each container can have several agents. A set of active agents is called a platform. Each JADE agent has a unique name for its identification. The main container in JADE has three automatically created agents called RMA: the Remote Management Agent which handles the GUI interface

AMS:
the Agent Management Service, which keeps of all agents and JADE programs DF: the Directory Facility, which advertises agent services The Main Container must be started first. A central Java RMI registry, keeps a record of all active containers to ensure effortless agent communication.  Agents were developed in SPADE and JADE and registered with the corresponding AMS. Messages were sent between agents. At first both platforms were hosted on the same system and then messages were sent between agents of the same platform. It was found that initially the processing time taken to complete the job was about the same. However, as the number of messages increased, JADE took slightly more time to complete the task. This was because messages were passed from one agent to the next only after a complete thread was formed between all the agents. Waiting time between agents increased slightly as none of the agents were ready to accept messages unless all of the agents were ready to accept messages. However, in SPADE it was found that as each agent completed its task, it became available for accepting the next task and so jobs were done asynchronously but were completed slightly faster. The rate of completion of jobs slightly increased with an increase in the number of messages. Fig. 9:-Rate of messages being processed It was also observed that as the messages being transferred increased, CPU performance was affected. The percentage of CPU usage in both JADE and SPADE increased. However, in JADE the percentage of CPU usage was comparatively lesser than in SPADE as the number of messages transmitted increased.

Conclusion:-
This paper has focused on the performance evaluation of co-operating multi agent systems on SPADE and JADE platforms after careful review with other agents. A multi-agent based system was also proposed to resolve security issues for agents developed using SPADE and JADE platforms. Both platforms identify network threats with the aid of agents, and makes suggestions to reduce them. SPADE agent system can utilise CPU optimally for communication and threat detection when compared to JADE. However, malware threat detection has not been implemented. The work will be proceeding further with the development of agents for malware threat detection using ML techniques.