WAVE: Popularity-based and Collaborative In-network Caching for Content-Oriented Networks K. D. Cho et al., IEEE INFOCOM 2012 Workshop, pp. 316-321, March 2012. January 17, 2013 Byeong-Gi Kim Park Laboratory, Waseda University
Outline Introduction WAVE: A Content Caching Scheme Simulation Results Conclusion
I. Introduction Content-Oriented Networking Content retrievals Efficient load balancing Content-oriented routing In-network Caching Faster content delivery Better content availability
I. Introduction (cont.) Chunk-based Caching Small sized chunks can provide shorter transfer delay and less processing overhead than file-based caching Replacing chunks instead of a whole file may increase the storage efficiency How to distribute chunks of the same file? Design Challenge Reduce communication and maintenance overhead E.g., web caches and CDNs require an explicit coordination between caches so substantial overhead is incurred Provide efficient content delivery and cache usage Lower latency, lower network resource use, etc.
II. WAVE: A Content Caching Scheme Characteristics Popularity-based Adjust the number of chunks to be cached considering the content popularity (i.e., access count) Simple Require no knowledge of access patterns Only two counters per file are required Decentralized No need of central server Incrementally deployable Downstream router can decide whether the router needs to cache the incoming data or not WAVE routers can operate with legacy routers
II. WAVE: A Content Caching Scheme Assumption Chunk Request instead of File Request 100 chunks will be requested by 100 chunk requests Chunk Marking and Forwarding Independent caching decision Minimum information for the caching decision is required Cache suggestion flag On-path Caching Chunk will be cached at a C-router along the path Chunk will be transferred from the C-router
II. WAVE (cont.) Fig. 1. Illustration of WAVE operations.
II. WAVE (cont.) Chunk Caching Algorithm What to cache What to replace Where to cache
II. WAVE (cont.) What to cache Fig. 2. Content caching example in a C-router.
II. WAVE (cont.) What to replace Use LRU (Least Recently Used) WAVE maintains the access history in the unit of a file to find a victim chunk to be replaced When the last cached chunk is replaced, the access history for the file can be removed
II. WAVE (cont.) Where to cache Direction and location should be considered carefully Direction WAVE distributes the chunks in the direction from which the chunk requests come Location One-hop distribution Multi-hop distribution ISP crossing distribution
III. Simulation Results Simulation Environment Use a discrete event-driven simulator Stub Stub Domain Domain (10) (10) Transit Domain (5) Stub Domain (10) Stub Domain (10) Stub Domain (10) 1,000 end hosts 10 original servers 100,000 files 1GB size content (Divided into 100 chunks) Zipf distribution with parameter 0.85 10GB storage size
III. Simulation Results (cont.) Network-wide Performance Average Hop Count Fast content retrieval Link Stress Traffic amount transferred over a particular link Load balance Inter-ISP Traffic Reduction The number of content downloads from original server
III. Simulation Results (cont.) Fig. 3. Performance comparison of WAVE against ProbCache, AllCache, CDN, and client-server.
III. Simulation Results (cont.) Cache-related Performance Cache Hit Ratio Cache Replacement Count Caching Efficiency Defined as the average number of cache hit counts divided by the number of aching events Relative Hop Count Defined as the ratio of how many hops a chunk is distant from the end hosts on average depending on its index to the average hop count Number of Chunks
III. Simulation Results (cont.) Fig. 4. Cache performance comparison of WAVE against ProbCache, UniCache, and AllCache model.
IV. Conclusion WAVE was proposed for efficient caching and delivery of content To reflect the content popularity, WAVE exponentially increases the number of chunks of a file to be cached as its request increases WAVE achieves higher cache hit ratio and less frequent cache replacements than other on-demand caching schemes