Asymptotic Analysis on Content Placement and Retrieval in mANeTs Jingjing Luo',Jinbei Zhangt,Ying Cuit,Li Yut,Xinbing Wang iSchool of Electronic Information and Communications,Huazhong University of Science and Technology,China Department of Electronic Engineering,Shanghai Jiao Tong University,China Email:fluojingjing,hustlyu}@hust.edu.cn,fabelchina,cuiying,xwang8}@sjtu.edu.cn Abstract-Recently,performance analysis for large-scale networks?Content-centric MANETs differ from traditional content-centric mobile ad hoc networks (MANETs)has received user-centric MANETs in the following two aspects.First,in intense attention.In content-centric MANETs,content delivery content-centric MANETs,content delivery is based on content consists of two operations,i.e.,content placement and content retrieval,which may involve different network costs.However, identifiers rather than locations of mobile users.Any user who existing performance studies in content-centric MANETs mainly has a content can act as a server for this content.Second. focus on content retrieval,and hence may not reflect the impact contents have heterogeneous popularity,which brings the need of content placement.In this paper,we investigate the asymptotic to design popularity-aware policies.Therefore,it is important throughput and delay performance by considering the two to understand how these new features fundamentally affect the operations of possibly different network costs.Specifically,we introduce a general weighted sum delay cost of content placement performance of content-centric MANETs. and content retrieval as the delay performance metric.We Recently,some initial work has already considered the consider an arbitrary content popularity distribution and study performance analysis of content-centric wireless networks. two mobility models in different time-scales,i.e.,fast and slow For example,in [13].Gitzenis et al.studied the required mobility.For each mobility model,we characterize the impacts link capacity for a static wireless network,where contents of the network parameters on the network performance.By optimizing the content placement and retrieval for contents of are requested according to a Zipf distribution [14].They different popularity,we design a general near-optimal scheme, demonstrated that caching contents can be beneficial for the the parameters of which reflect the delay weights of the two sustainability of networks expanding in size.This promising phases.We show that the network performance improves as the result suggests that CCN may break the bottleneck of wire- number of cached replicas increases until the number reaches a threshold.Finally,we show that our results are general and can less communications,as the traditional user-centric network incorporate some existing results as special cases. is shown to be non-scalable [15].In [161.the asymptotic throughput of content-centric wireless networks with given content lifetime was studied.It was shown that increasing the I.INTRODUCTION content lifetime with the size of the network may result in Nowadays,instead of being concerned about communi- higher throughput.On the other hand,the impact of mobility cating with specific hosts,users are mainly interested in on content-centric wireless networks was investigated in [17]. obtaining contents they desire.To reflect this change,content- In particular,it was assumed that nodes independently and centric networking (CCN)[1]-[3]is proposed as a promising uniformly visit the network and the cache size of each node architecture for future Internet,which enables efficient content for storing contents is limited.Under this assumption,it was delivery based on content identifiers rather than host address- shown that the best throughput-delay tradeoff is achieved in es.This emerging CCN is currently changing the landscape the quasi-static case and mobility has a negative impact on the of the research for wireline networks.As direct device-to- network performance.This is quite counter-intuitive,as it is device (D2D)data sharing among mobile users is increasingly well known that mobility can improve the performance of the popular,CCN is also showing great potentials in designing traditional user-centric MANETs [61. infrastructure-less mobile environments like MANETs [4].[5]. In content-centric MANETs.content delivery consists of The performance investigation of content-centric MANETs is two operations,i.e.,content placement and content retrieval. therefore of great interest. These two operations may proceed concurrently or at different In traditional user-centric MANETs,the mobility of users periods during which the traffic load may vary significantly, can bring a dramatic improvement in throughput since mobile and hence result in different consumptions of network re- users can store packets and physically carry them while sources.For example,units of transmission costs in peak- moving around the network [6].This improvement comes hour and off-peak hour may be different.However,in wireless at the cost of an excessive delay.Therefore,there exists a CCNs,placement and retrieval are usually considered sepa- tradeoff between throughput and delay,which has been widely rately [13][16][17].Furthermore,in [13][16][17],for ease studied in related work on user-centric MANETs [7]-[12].of analysis,placement is assumed to be given and free,and An interesting question then is:can content-centric MANETs only retrieval cost is considered in the performance analysis. further improve the tradeoff by taking advantage of both These models may be valid when mainly focusing on content user mobility and key features of content-centric wireless retrieval.However,since placement also consumes network1 Asymptotic Analysis on Content Placement and Retrieval in MANETs Jingjing Luo† , Jinbei Zhang‡ , Ying Cui‡ , Li Yu† , Xinbing Wang‡ †School of Electronic Information and Communications, Huazhong University of Science and Technology, China ‡Department of Electronic Engineering, Shanghai Jiao Tong University, China Email: †{luojingjing, hustlyu}@hust.edu.cn, ‡{abelchina, cuiying, xwang8}@sjtu.edu.cn Abstract—Recently, performance analysis for large-scale content-centric mobile ad hoc networks (MANETs) has received intense attention. In content-centric MANETs, content delivery consists of two operations, i.e., content placement and content retrieval, which may involve different network costs. However, existing performance studies in content-centric MANETs mainly focus on content retrieval, and hence may not reflect the impact of content placement. In this paper, we investigate the asymptotic throughput and delay performance by considering the two operations of possibly different network costs. Specifically, we introduce a general weighted sum delay cost of content placement and content retrieval as the delay performance metric. We consider an arbitrary content popularity distribution and study two mobility models in different time-scales, i.e., fast and slow mobility. For each mobility model, we characterize the impacts of the network parameters on the network performance. By optimizing the content placement and retrieval for contents of different popularity, we design a general near-optimal scheme, the parameters of which reflect the delay weights of the two phases. We show that the network performance improves as the number of cached replicas increases until the number reaches a threshold. Finally, we show that our results are general and can incorporate some existing results as special cases. I. INTRODUCTION Nowadays, instead of being concerned about communi￾cating with specific hosts, users are mainly interested in obtaining contents they desire. To reflect this change, content￾centric networking (CCN) [1]–[3] is proposed as a promising architecture for future Internet, which enables efficient content delivery based on content identifiers rather than host address￾es. This emerging CCN is currently changing the landscape of the research for wireline networks. As direct device-to￾device (D2D) data sharing among mobile users is increasingly popular, CCN is also showing great potentials in designing infrastructure-less mobile environments like MANETs [4], [5]. The performance investigation of content-centric MANETs is therefore of great interest. In traditional user-centric MANETs, the mobility of users can bring a dramatic improvement in throughput since mobile users can store packets and physically carry them while moving around the network [6]. This improvement comes at the cost of an excessive delay. Therefore, there exists a tradeoff between throughput and delay, which has been widely studied in related work on user-centric MANETs [7]–[12]. An interesting question then is: can content-centric MANETs further improve the tradeoff by taking advantage of both user mobility and key features of content-centric wireless networks? Content-centric MANETs differ from traditional user-centric MANETs in the following two aspects. First, in content-centric MANETs, content delivery is based on content identifiers rather than locations of mobile users. Any user who has a content can act as a server for this content. Second, contents have heterogeneous popularity, which brings the need to design popularity-aware policies. Therefore, it is important to understand how these new features fundamentally affect the performance of content-centric MANETs. Recently, some initial work has already considered the performance analysis of content-centric wireless networks. For example, in [13], Gitzenis et al. studied the required link capacity for a static wireless network, where contents are requested according to a Zipf distribution [14]. They demonstrated that caching contents can be beneficial for the sustainability of networks expanding in size. This promising result suggests that CCN may break the bottleneck of wire￾less communications, as the traditional user-centric network is shown to be non-scalable [15]. In [16], the asymptotic throughput of content-centric wireless networks with given content lifetime was studied. It was shown that increasing the content lifetime with the size of the network may result in higher throughput. On the other hand, the impact of mobility on content-centric wireless networks was investigated in [17]. In particular, it was assumed that nodes independently and uniformly visit the network and the cache size of each node for storing contents is limited. Under this assumption, it was shown that the best throughput-delay tradeoff is achieved in the quasi-static case and mobility has a negative impact on the network performance. This is quite counter-intuitive, as it is well known that mobility can improve the performance of the traditional user-centric MANETs [6]. In content-centric MANETs, content delivery consists of two operations, i.e., content placement and content retrieval. These two operations may proceed concurrently or at different periods during which the traffic load may vary significantly, and hence result in different consumptions of network re￾sources. For example, units of transmission costs in peak￾hour and off-peak hour may be different. However, in wireless CCNs, placement and retrieval are usually considered sepa￾rately [13] [16] [17]. Furthermore, in [13] [16] [17], for ease of analysis, placement is assumed to be given and free, and only retrieval cost is considered in the performance analysis. These models may be valid when mainly focusing on content retrieval. However, since placement also consumes network