Summary by Patrick This paper describes a probabilistic peer-to-peer publish-subscribe system that does not utilize rendezvous nodes and is scaleable. Group members install group-specific routing vectors in neighbours up to a few hops away in order to act as a gravity well. Each node shares which groups it is a member of with its overlay nodes. Messages are sent by two methods in the network. Either randomly or to a specific node depending on wether or not the node is a member of a group. Evaluation is done by simulation based on random networks of a fixed size and based on data from MySpace. The results show a high scaleability and decent coverage of the network. Suggested Questions: * Is the system vulnerable to attack by dishonest peers? * Could this system be applied to other areas, like bittorrent? Summary by Zahid As compared to the old way of Peer to peer publish subscription depending on structured and avenues nodes which are stored and depend on group membership¡¯s information. Quasar is rendezvous less subscriber system that provides the specific needs of social networks and its managed the different social network¡¯s group members. So these members are managed in this system. This paper highlight the importance of Quasar system that focus on probabilistic publish subscribe system for social networks. In this paper authors described the suggested system Quasar having a scale, peer-to-peer publish subscribe system that provides the specific needs of a social network. Quasar introduce a new way through which a combination of proactive dissemination of aggregated routing vectors those help to provide an efficient any cast primitive to figure out the group members whose are near to each others. Summary by Adnan Introduction : Social media sites like facebook has more traffic today than the typical emailing systems. Thus the number of users are increasing and so is the traffic. At backend a centralized system is maintained but that concerned by the end user in terms of privacy. It also leads to issues like poor scalability. This paper outlines Quasar, a scalable, peer?to? peer publish?subscribe system that caters to the specific needs of social networks. The number of users in a network is ???? considered and these users or nodes maintain state that is constant in the number of overlay links.So on the basis of these overlay network the topology is decided. Anycast routing mechanism is used in quasar which is different the the anycast IP routing. The node generating the traffic does not get back the traffic in quasar this helps to avoid the unnecessary generation of loops. Overall, this paper makes three contributions. It introduces a scale?free, anycast?like, rendezvous?less routing primitive for overlay networks. Second, it describes a novel publish?subscribe system that explores a new point in the design space with benefits that cater to the requirements of social?networks. Finally, it shows the system is practical, efficient and hotspot?free in both a large?scale simulation consisting of a social network of 32,000 nodes, and a smaller simulation parameterized with MySpace profile data of over 10,000 users¡¯ friends lists and communities. Quasar, a scalable, peer?to? peer publish?subscribe system that uses different techniques to cover need sof different social networks.To find near by group memebers it uses two machnisum togeather to make it efficient. Bother aggregated routing and directed walks. Multicast topology sits on top of the previously described anycast topology thus helping to find more nodes efficiently Summary by Ioana Quasar: a probabilistic publish-subscribe system for social networks - Bernard Wong and Saikat Guha (IPTPS 2008) The paper introduces peer-to-peer publish-subscribe system for social networks called Quasar. The contributions of the paper are : the scale-free, anycast- like , rendezvous-less event routing primitives for overlay networks, the description of the new publish-subscribe system and the evaluation of the proposed system in large-scale simulation ( ~ 32.000 nodes ) and in smaller simlation parametrised with MySpace profiles( ~ 10.000 users ) The description of the system( Section 3 ) presents the joining protocol of a new user to a group and the routing protocol used to publish a message to a group. In the proposed system, each node represents a user of the social network. Multiple groups exist within the network. Groups and nodes are identified based on identification code ( which represents a 160-bit hash of their names). The process of a user joining to a new group involves subscriptions dissiminating proactively to a fixed depth. The information that is dissimiated is is represented by attenuated bloom filters. Routing a message to group is a process in which the message can be routed to multiple groups before it reaches the destination. The message is routed by a random walk. At each node the matching group ID is checked. If the message reached the destination group it is delivered, otherwise the bloom filter of the node is used to check whether there is a match for the group ID. The message is routed until a match is found or the TTL of the messsge expires. The routing process takes into account the negative informations. The authors evaluate their approch by using a large simulation ( up to 32768 nodes ) parametrized by MySpace data. For the the total number of profiles considered ( 10.930 ), the friends lists contain, on avarage, 916 users ( but a very small fraction of these are online ). Two metrics are used to evuate the system : coverage and performance . The coverage, as it is defined in the paper, refers to the fraction of members of group that reveive a published message. And performace referes to the latency ( the upper bound ) between publishing a message and the last group member that reveives the message. For the MySpace sample data, both of the metrics show promising results : 95% for the coverage and, a very small increase in the performance there is an increase in the number of nodes. Questions : 1. What other metrics can be used to evaluate the system ?