Research On Content Caching And Delivery In Wireless Networks With Client Pre-caching

The efficient dissemination of multimedia content has been one of the key means to tackle the massive data challenge.Unlike voice services,multimedia content has its unique diversity property,i.e.,most requests concentrate on a small number of popular contents.Also,a majority of requests arrive asynchronously.On the other hand,caching storage at mobile edge has maintained a steady growth.In order to efficiently deliver multimedia content under asynchronous requests,we utilize both the content diversity property and the ever-increasing client cache storage,focusing on the following three key technologies: a)Before request arrival times,content pushing and client pre-caching are adopted to satisfy all asynchronous requests for the cached hot videos in advance via a single broadcast pushing;b)At request arrival times,multicast delivery and client pre-buffering could be used to take advantage of the multicast opportunities among adjacent requests for hot videos;c)As for nearby clients,Device-to-Device(D2D)transmission and cooperative caching could be adopted to further utilize the D2 D offloading opportunities.Based on these three ideas and step by step,we study the caching and delivery design of three different scenarios,i.e.,a broadcast-cellular converged network with content pushing and limited storage,video-on-demand(Vo D)services with client pre-caching and multicast delivery,Vo D services with D2 D offloading and base station(BS)multicast.These analyses serve as a theoretical basis and guidance for caching and delivery design in wireless networks with client pre-caching.The main contributions of this paper are as follows:· Network capacity analysis and pushing and caching design for the broadcastcellular converged network with limited cache storage: In order to reduce the redundant transmission of hot contents under asynchronous requests,a broadcast-cellular converged network with content pushing and client pre-caching is constructed,where the broadcast network pushes hot contents into the limited cache storage at client side.Compared with previous work based on unlimited storage assumption,we investigate the network capacity of the converged network with limited cache storage,and even more,under the general storage distribution for the first time.Hence the considered scenario is more realistic and the results are more general.Based on the different relationship between the pushing capacity and the caching ability,the system can be further divided into different operational regions,which reveal the constraining factor on the growth of network capacity.Moreover,the optimal pushing and caching schemes are investigated from both the network and client perspectives.It is shown that both criteria reduce to the same offloading objective.Based on different traffic loads and operational regions,the offloading optimization problem yields to different pushing and caching schemes,where the traditional popularity-based pushing and caching scheme performs the best in the cache-limited region under low traffic load.These analyses provide guidance on improving the network capacity and designing the pushing and caching scheme in the converged network with limited storage.· Joint design of caching and delivery for VoD services in reactive and proactive modes: In order to further utilize the multicast opportunities of hot videos at request arrival times,we jointly optimize the cache allocation and the delivery scheme of Vo D services in both reactive and proactive modes based on multicast delivery and client prebuffering.In addition to the traditional full-access pattern where clients watch the desired video entirely,this thesis also considers the caching and delivery design of intervalaccess patterns for the first time.In the reactive mode where videos are served on demand,we jointly design the cache allocation and multicast delivery in order to minimize the average bandwidth consumption for Vo D services with zero-delay constraint.Specifically,a client caching enabled stream merging(CCE-SM)scheme is developed for multicast delivery and the optimality is proved.For the first time,the minimum bandwidth consumption for Vo D services with client pre-caching under Poisson arrivals is obtained.The cache allocation problems under different access patterns are proved to be convex optimization problems or fractional knapsack problems,where the close-form solution can be obtained.Then,in the proactive mode where the videos are periodically broadcast,we jointly design the cache-bandwidth allocation and the multicast delivery to minimize the average waiting time under the total bandwidth constraint.These analyses also reveal the trade-off between bandwidth and cache resource for a certain client request rate and delay requirement.· Cache allocation and multicast delivery for Vo D services with client pre-caching and adaptive multicast: In order to utilize the channel difference among heterogeneous clients in the wireless scenario,we apply the adaptive modulation and coding to the stream merging technique for the first time,and propose an adaptive-multicast streammerging scheme with client pre-caching.Based on the reward renewal theorem,we derive the bandwidth consumption of the proposed delivery scheme for a single video under a given cache allocation.Then the cache allocation problem to minimize the total bandwidth consumption for multi-videos is proved to be convex optimization problems.Results show that under relatively low client request rate,the proposed scheme reduces to the unicast delivery and the optimal cache allocation is to cache the most popular contents entirely;Under high request rate,the proposed scheme reduces to the broadcast delivery eventually and the adaptive multicast gain vanishes,where the optimal cache allocation can be obtained via the water-filling method;Under relatively high request rate,the proposed scheme further becomes periodically broadcasting and the bandwidth consumption is fixed.Compared to existing schemes,the proposed scheme fully exploits the content diversity among asynchronous requests from clients with different channel conditions,yielding a great reduction in the bandwidth consumption.· Caching and delivery design for VoD services with D2 D communication and BS multicast: In order to further utilize the potential D2 D offloading opportunities among nearby clients,we propose a delivery scheme which combines cooperative caching and BS multicast.A requesting client can obtain the desired content from its local cache,other clients it meets via out-of-band D2 D communication within a maximum tolerant delay,and the BS multicast transmission.We optimize the caching allocation and the delivery method to minimize the BS multicast traffic.The cache allocation problem can be modeled as a separable concave minimization problem with convex constraints,which can be solved via the convex underestimation and branch bound,where the global optimality is achieved at the cost of a very high computational complexity.To reduce the complexity,we further adopt the convex-concave procedure(CCP)to obtain a stationary point.Compared to the existing work which separately focuses on multicast delivery and D2 D offloading,the proposed scheme utilizes both effectively,and achieves a significant performance improvement.