3. (20 pts) Consider a decision-making about the placement of mirror sites in a content distribution network (e.g., NetFlix video cache sites). See Figure 2 for illustration. The decision-making is a fuzzy process, due to the following unclear trade-offs: i) Faster content access by clients vis-a-vis higher overhead of updating mirror sites; ii) Higher monetary cost involved in deploying a large number of mirror sites; iii) Security risks from adversaries due to wider exposure of content with higher replication; iv) Increased software complexity of running push/pull algorithm on many mirror sites. Discuss the above trade-offs (i)-(iv) in a semi-quantitative manner. You can use, for instance, different client-level content demand rates, content-update rates (e.g., frequency of news postings), economic aspects of leasing machines from data-centers, network bandwidth allocation costs, and customer dis- satisfaction due to content download delays. content updates [e.g., news, events, video/images, software patches, . .] End-to-end algorithms Pull: client reaches server site to read content pages N Push: server reaches client site to deliver content pages content-data cache site transit node capable to store and/or forward content-data over network link pull(Pa) cl Pa data-center in Philadelphia customers in New York City push(Pc) data-center in Tokyo customers in I South-Asian region Pb c2 Pc QoS: quality of content delivery service CONTENT server (R) (processes contents and stores content pages) data-center in Dallas customers in SE-Asian region c3 INTERMEDIATE PUSH-PULL boundary content relay network (say, set up over Internet) CN Texas, & Arizona customers client terminals (E-readers, smart-phones, pocket-PCs, ..) Lower latency; local feed from cache site via network attachment Reduced bandwidth/storage consumption; to effect content-data delivery to client devices Figure 2: Illustration of a content-delivery network infrastructure (e.g., Net Flix, YouTube)