580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can clustering web content optimize replication strategies to improve client latency and reduce replication overhead in CDNs?
This research theme investigates the role of clustering techniques in grouping web content for efficient replication in Content Distribution Networks (CDNs). It addresses how clustering can balance the tradeoff between fine-grained replication (e.g., per URL) and coarse-grained replication (e.g., per website), targeting reductions in client latency and minimizing replication and management costs. The significance lies in enabling scalable, adaptive content distribution that maintains performance while reducing network and computational overhead.
Key finding: The study demonstrates that cooperative pushing of web content to CDN nodes can achieve comparable user-perceived latency with only 4-5% of the replication and update traffic compared to uncooperative pulling approaches.... Read more
Key finding: Extending prior work, this paper confirms that clustering web content based on request correlations enables efficient replication strategies that balance performance and overhead. Offline clustering based on historical access... Read more
Key finding: This work provides a systematic comparative study of different similarity measures (Euclidean, cosine, Pearson correlation, extended Jaccard) coupled with multiple clustering algorithms (self-organizing feature map,... Read more
Key finding: Through an analytical survey, this paper highlights essential requirements for web document clustering such as relevance, browsable summaries, overlap handling, snippet tolerance, speed, and incrementality to effectively... Read more
Key finding: By providing a comprehensive overview of clustering methods relevant to web data, including classical, graph-based, and neural network techniques, this paper contextualizes clustering approaches in the web environment. It... Read more
2. What are the challenges and strategies for adaptive and incremental clustering to maintain replication efficiency amid changing web access patterns?
This theme covers research focusing on how CDNs can dynamically update their content clustering and replication strategies as user access patterns evolve. It addresses maintaining cluster relevance over time to sustain performance benefits while minimizing overhead in reclustering and redistribution. Adaptivity in clustering ensures replication strategies remain effective in volatile web environments, especially during flash crowds or rapid content shifts.
Key finding: The paper identifies that clustering based on outdated access patterns degrades replication performance beyond one week. It shows that complete reclustering improves performance but incurs prohibitive overhead. To balance... Read more
Key finding: This study explores both offline and online incremental clustering for adapting replication to evolving user accesses. Offline methods using prior access history achieve near-complete reclustering performance with less... Read more
Key finding: This work introduces the concept of ephemeral clustering, where document sets are dynamically generated (e.g. from search results) and clusters have a short lifespan for interactive browsing. It highlights that ephemeral... Read more
Key finding: By evaluating different similarity measures and clustering algorithms for high-dimensional sparse web data, this paper provides tools crucial for designing adaptive clustering systems that maintain quality over time. The... Read more
3. How do underlying network mechanisms and architectures influence web content replication, and what role do clustering and replication placement algorithms play in optimizing content delivery?
This theme synthesizes insights on how network-level designs, including replica server placement, network coding, and security considerations, impact web content replication efficiency and reliability. It explores algorithms for replica placement within Content Delivery Networks (CDNs), innovations in data dissemination leveraging network coding, and security architectures ensuring integrity in replicated environments, elucidating the broader system context into which clustering and replication strategies are embedded.
Key finding: This comprehensive survey categorizes and compares replica server placement algorithms in traditional and emerging CDNs architectures, including cloud and NFV-based CDNs. The paper identifies key requirements such as cost... Read more
Key finding: The research introduces network coding at intermediate nodes as a method to enhance content distribution efficiency in large, dynamic networks. Simulations show network coding improves expected file download time by more than... Read more
Key finding: This work presents a unified object model integrating data content, replication strategies, and security architecture to provide integrity guarantees for replicated Web documents on untrusted servers, including CDN nodes. It... Read more
Key finding: The thesis provides analytical models and experimental insights for scalable content distribution involving caching, multicast, and their combinations. It evaluates various performance parameters such as latency, server load,... Read more
Key finding: By analyzing transformations of web request streams through aggregation, disaggregation, and filtering by web components, this paper elucidates how temporal locality properties evolve in the Web. It explains the impact of... Read more