**An Analysis of Caching in Information-Centric Vehicular Networks

Detailed

1、In the case of traffic data dissemination, location and temporal requirements need to be evaluated both in terms of data validity, usefulness as well as on where this content is relevant.
2、content copies should be placed where they are most likely to be requested
3、lookup time(查询时间)+cache size(缓存大小)
4、仿真：The ICN-VANETs In-network caching models were incorporated into the Veins simulation framework (veins.car2x.org), OMNeT++ network simulator
(omnetpp.org), and SUMO road traffic simulator (sumo.dlr.de)

思考：

1、content应该被分为两部分，

• Fir. 全局流行内容。这种content在整个网络中被请求次数最多，每个RSU应该都有备份；
• Sec. 局部流行内容(location dependent content)。这种内容呈现地域性。比如，北京人关注北京事；交通信息。再细分的话，由于车的流动性，流行的数据在不同地区也呈现相似性。

2、分析了cache的种类，但是相对笼统，举出的参考文献较旧，偏向纯caching，contribution价值不大。

• Implicit Caching Coodination:[2]
• Explicit Caching Coordination:[5]
  1、Cache policies are maintained by the BSs and can be coordinated with RSUs to predictably cache content a given vehicle behavior.
• Chunk-level Caching:[6]
  1、Universal naming allows for partial content storage which can further reduce communication overheads and can be incorporated into both implicit and explicit caching policies.
  2、Chunk storage efficiency depends heavily on content granularity and correlation. The access frequency of individual chunks will differ and similarly to objects, content chunks follow will follow some distribution, very commonly it being the Zipf distribution [6]. It should be noted that file popularity cannot be extended to the chunk level as they follow independent distributions.
  3、Hence, implementing chunk-level caching requires additional knowledge, harder to maintain in systems such as ICN-VANETs and whose implementation poses additional challenges.
• Predictive Caching:[7]
  1、Predictive caching takes advantage of network knowledge to improve content delivery
• Mobility-aware Caching:[8]