Support for Peer-to-Peer Interactions in Web Brokering Systems.Geoffrey Fox and Shrideep PallickaraCommunity Grid LabsDepartment of Computer ScienceIndiana University.The peer-to-peer (abbreviated as P2P) style interaction model is a very powerful one and is one in which clients caninteract directly with each other. The traditional boundaries between clients and servers are blurred with clientsinitiating sophisticated requests and those requests being serviced by other clients. The P2P style model facilitatessophisticated resource sharing environments. Driven entirely on demand a resource may be replicated several times;a process that is decentralized and one over which the original peer that advertised the resource has sometimes littlecontrol over. The greater the demand the greater the system initiated replications for the resulting in fewerbottlenecks and faster accesses to that resource. Peers usually do their interactions using XML, this ensures thatpeers could be written in any language and be compiled for any platform, and thus interactions between peers couldbe cross-platform and cross-language.The p2p request/response semantics differ from those prevalent in traditional systems, where the request/responsemodel is fixed (one response for a request) with the client initiating the request having no ambiguity on how therequest would be interpreted and also being aware of what the response would be like. In contrast in p2p systemsnot every request goes through and even if it does there could be zero or more valid responses and peers anticipateneither the template that the responses would conform to nor on the order in which these responses would bereceived. Furthermore, responses are not identical with each responding peer processing the request based on theresources at its disposal and its interpretation of the request. Requesting peers and those peers that bounce therequests/responses are thus made aware of different capabilities that exist at other peers. This discovery of servicesoffered by other peers constitutes dynamic real time knowledge propagation. Peer requests are sometimes satisfiedthrough cached responses, and peers generally have a choice on whether to accept these cached responses or not. Itsup to the peer to discard responses that it deems is not right. P2P interactions are self-attenuating with requests dying out after a certain number of hops. These attenuations intandem with traces of the peers that the interactions have passed through eliminate the continuous echoing problemthat result from loops in peer connectivity. However attenuation of interactions sometimes prevents clients fromdiscovering certain services that are being offered. Peers hosting these services could not respond because themessage attenuated before it could have reached the hosting peer. This results in p2p interactions being verylocalized. Of course if the peer at the edge of the attenuation had cached the response for a similar request that itmade or the resource itself that particular request would be available to the requesting peer. Resources in web brokering systems are generally within the purview of the broker network. P2P systems comprisepeers exposing the resources they share. Unlike clients in web brokering systems that interact via the brokernetwork, peers in p2p systems interact directly with each other and sometimes use other peers as intermediaries ininteractions. Specialized peers are sometimes deployed to enhance routing characteristics. Nevertheless,sophisticated routing schemes are seldom in place and interactions are primarily through simple forwarding ofrequests with the propagation range being determined by the attenuation indicated in the message. This is where distributed web brokering systems could come in place. Having a single broker solution would lead tobottlenecks where a lot of p2p interactions are being funneled through the broker. The associated queuing delays,scaling issues and the single point of failure that such a scheme constitutes are among the reasons why a distributedmodel should be in place. Distributed brokering systems could be used to optimize the request/response/discoveryand advertisement interactions. Furthermore these systems could be used to connect islands of peers together. Peersthat are not directly connected through the peer network could be indirectly connected through the broker network. Integration strategies need to ensure that minimal changes need to be made to the brokering core, zero changes tothe peers and no straitjacketing of any interactions that peers had prior to integration. This would be done viaproxies that provide an interconnection bridge between the two systems. The proxies are part routing peers and partclients of the web brokering system. Peers would interact with the proxies as it would with any other peer, while theproxy also inherits guarantees accorded to clients of the brokering system. Thus no peers need to be changed, neitherwould this entail any major support within the existing distributed broker network. The broker network would beused primarily as a delivery engine and a pretty efficient one at that while locating peers and propagatinginteractions to relevant peers. Broker networks are also best suited to react to changes in peer requests, concentrations and resource availability.Brokers/links can be dynamically instantiated or purged to compensate such changing conditions. Peers may beimplemented in different languages with interoperability being achieved through XML-based data interchange.XML’s data description and encapsulation properties allow for ease of accessing specific elements of data that isthen used to achieve best possible routing characteristics. Similarly some resources are best managed by the systemrather than being left to the discretion of peers who may or may not be present at any given time. An understandingof the network topology and an ability to pin point the existence of peers interested in that resource are paramount toefficient replications of a resource. The distributed broker network best handles this.In supporting p2p interactions the broker network itself should not be flooded with the processing of duplicatemessages as resulting from message propagations from multiple peers as a result of loops in peer connectivity.