580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can peer-to-peer architectures enhance scalability and resource sharing in massively distributed collaboration systems?
This research theme focuses on leveraging peer-to-peer (P2P) communication and computing paradigms to improve scalability, resource utilization, and robustness in large-scale collaborative networks. It examines how decentralized architectures can overcome bottlenecks typical in centralized client-server models, enabling cooperative sharing of resources such as computing power, storage, and bandwidth among numerous participants. This line of research is critical to supporting massive collaborations where centralized coordination is infeasible and where participants dynamically join and leave.
Key finding: The ShareGrid project demonstrated that federating computing resources from small independent research labs using a P2P desktop grid is feasible and effective, enabling consortiums of smaller organizations to collectively... Read more
Using Peer-to-Peer Communication to Improve the Performance of Distributed Computing on the Internet
Key finding: This work established that introducing P2P communication to distributed computing platforms significantly reduces the bandwidth bottleneck on centralized schedulers by enabling direct data transfer among participants.... Read more
Key finding: The survey demonstrated that collaboration between content delivery systems and network infrastructure, including the incorporation of P2P mechanisms, can enhance content distribution efficiency by exploiting network topology... Read more
Key finding: This paper proposed a P2P-based architecture for open-source software distribution addressing scalability, load balancing, and robustness challenges in large dynamic communities. It integrated distributed metadata indexing... Read more
Key finding: Introducing network coding into P2P content distribution leads to significantly better expected file download times—up to 20-30% improvement compared with coding at the server only, and more than two-fold compared to sending... Read more
2. What models and mechanisms explain and predict user activity dynamics and information sharing in massively collaborative online networks?
Understanding how user activity evolves and information propagates within large online collaboration networks is fundamental for designing self-sustaining systems. This theme investigates mathematical and computational models capturing peer influence, activity decay, random and strategic information forwarding, and network topologies (small-world, scale-free). These insights help identify critical factors driving collaboration longevity, information dissemination efficiency, and strategies to overcome participation and contribution challenges in massive, distributed user communities.
Key finding: This study formulated a continuous dynamical system combining intrinsic activity decay and positive peer influence to model user activity in online collaboration networks. It experimentally showed that activity dynamics have... Read more
Key finding: Through analytical bounds and empirical evaluation, this paper demonstrated that random forwarding of information in collaboration networks performs surprisingly close to optimal sharing strategies, especially in small-world... Read more
Key finding: By analyzing the social network of creators producing videos on the Japanese site Nico Nico Douga, the study revealed that different creator categories fulfill distinct roles within an emergent massive collaborative... Read more
Key finding: While primarily a theoretical work, this paper identified information dissemination and coordination in networks—e.g., via oblivious routing and consensus problems—as key challenges in distributed systems, which underpin... Read more
Key finding: The VAST architecture introduced a spatial publish/subscribe model utilizing a peer-to-peer Voronoi overlay network to enable scalable, efficient, and self-organizing message dissemination in large collaborative networks. The... Read more
3. What data structures and protocols enable efficient consistency, ordering, and coordination in massively distributed collaborative editing and multi-agent systems?
This theme explores methods for maintaining consistent and coherent shared states, such as documents or multi-agent plans, in massively distributed settings characterized by concurrent user actions and partial knowledge. It addresses sequence CRDTs for ordered collaborative editing, flexible sub-team coordination in large agent teams, and decentralized user modeling. The goal is to develop scalable, fault-tolerant, and adaptive coordination mechanisms that preserve intention and enable real-time collaboration without centralized control.
Key finding: LSEQ proposed an adaptive allocation strategy for sequence CRDTs in distributed collaborative editing to reduce the typically linear growth of identifier size to sub-linear on average. This optimization improves space... Read more
Key finding: The authors introduced a fully decentralized user modeling architecture where each peer independently extracts, translates, and communicates user profile data using predefined templates via an internal broker. This P2P design... Read more
Key finding: This research developed an algorithmic framework organizing large multi-agent teams into dynamically evolving overlapping subteams connected by a decentralized social network topology with small-world properties. By... Read more
Key finding: NaradaBrokering, a distributed messaging middleware based on publish/subscribe, was demonstrated as capable of supporting scalable, fault-tolerant, and globally distributed collaborations over inter-continental networks. Its... Read more
Key finding: DecentralizePy is a modular distributed framework designed to emulate large-scale decentralized machine learning over arbitrary topologies. It enables studying topology impacts, communication-efficient aggregation techniques... Read more