Optimizing Caching Policy for Loss Recovery in Reliable Multicast

Video caching at the wireless edge is a promising approach to improve throughput and reliability of video delivery. While most existing literature considers homogeneous user preference modeling when designing caching policies, this paper investigates how to exploit statistical information on individual popularity preferences in the design of caching policies in base station assissted wireless device-to-device networks. Specifically, we formulate the caching policy design problem aiming to minimize average energy consumption and propose two approaches that solve the problem under different conditions: users can or cannot coordinatedly design their policies. By simulating systems using practical individual preference models, we show that the proposed designs are effective, and the performance gain brought by exploiting individual preferences can be clearly observed.

Multicasting plays an important role in internet networks. Supporting wavelength division multiplexed (WDM) multicasting in an IP over WDM network poses interesting problems because some WDM switches may be incapable of switching an incoming signal to more than one output interface. Although some multicast applications, such as audio/video conferencing can tolerate packet error or loss, other applications such as one-to-many file transfer and military applications do not. Hence, an inevitable need for reliable multicasting arises. The reliable multicast protocol ensures the eventual delivery of multicast data to all multicast group members. The reliable multicast is receiver initiated NACK-based and it makes use of feedback suppression in order to avoid NACK and retransmission implosion. The loss recovery burden is distributed over the multicast group members in order to reduce recovery latency and end-to-end delay. The proposed protocol is evaluated in terms of performance metrics for an IP/WDM optical network which will be the candidate for the next generation internet networks.

Proactive edge caching, as a promising approach to accommodate the explosively increased mobile data demand of Video on demand (VoD) service, has received extensive attention. However, although targeted to VoD service, existing caching policies are mainly designed for file downloading service while the unique requirement of quality of experience (QoE) for VoD service has been seldom considered. Aimed at maximizing the weighted average QoE of VoD service, this paper optimizes the proactive edge caching polices including the cached fraction and encoding bit rate of every video. We consider a two-tier network, where the helpers equipped with caching resource are deployed in the coverage area of traditional base stations. We formulate the caching optimization problems and show their hidden convexity properties, then we find that the optimal caching policy is determined by the weighted popularity-to-duration ratio of videos. Based on the result, we develop a low-complexity algorithm to find the optimal caching policy. Simulation results demonstrate evident performance gain of the proposed policy over the existing policy for VoD service.

Local loss recovery for reliable multicast can provide significant performance improvement in terms of loss recovery latency, bandwidth consumption and network throughput. Active reliable multicast (ARM) is a novel loss recovery scheme for large-scale reliable multicast. When sufficient active resources of active routers (ARs) and associated cache are available, ARM gives good loss recovery latency performance. The authors build an analytical model for studying the optimal active resource allocation for ARM. They propose a top-down layering analysis (TDLA) approach to derive the expected loss recovery latency (ELRL). For a given multicast tree, the ELRL is derived as a function of the active resource allocation pattern. Simulation experiments reveal that high performance gains can be attained with limited active resources, which are properly allocated. It is also shown that simulation results have a good match with the analytical results. The analytical framework developed can be used for further investigation of various active resource allocation schemes and caching policies.

Packet loss recovery is a key issue in reliable multicast. An effective way for packet loss recovery is to place repair servers with active routers along the transmission paths. These repair servers naturally form a hierarchical cache system due to the hierarchical nature of the multicast tree. How to design an effective hierarchical cache system to minimize the packet loss is important. In this paper, we first derive a cooperative caching efficiency model for a hierarchical cache system. Based on the model, a heuristic Cooperative Cache Replacement (CCR) algorithm is proposed to achieve efficient cache performance for reliable multicast systems. The implementation issues are also discussed in detail. The ns-2 based simulations are conducted to evaluate the performance of the proposed algorithm by compared to the optimal caching time (OCT) based algorithm. The results show that CCR effectively reduces the packet loss recovery latency.

Most computer systems currently consist of DRAM as main memory and hard disk drives (HDDs) as storage devices. Due to the volatile nature of DRAM, the main memory may suffer from data loss in the event of power failures or system crashes. With rapid development of new types of non-volatile memory (NVRAM), such as PCM, Memristor, and STT-RAM, it becomes likely that one of these technologies will replace DRAM as main memory in the not-too-distant future. In an NVRAM based buffer cache, any updated pages can be kept longer without the urgency to be flushed to HDDs. This opens opportunities for designing new buffer cache policies that can achieve better storage performance. However, it is challenging to design a policy that can also increase the cache hit ratio. In this paper, we propose a buffer cache policy, named I/O-Cache, that regroups and synchronizes long sets of consecutive dirty pages to take advantage of HDDs' fast sequential access speed and the non-volatile property of NVRAM. In addition, our new policy can dynamically separate the whole cache into a dirty cache and a clean cache, according to the characteristics of the workload, to decrease storage writes. We evaluate our scheme with various traces. The experimental results show that I/O-Cache shortens I/O completion time, decreases the number of I/O requests, and improves the cache hit ratio compared with existing cache policies.

Content caching in vehicular networks is a promising technology to dramatically reduce the request-response time and transmission delay. The existing caching policies often suffer from high computation and communication overhead and fail to well capture the dynamics of the vehicular networks and content popularity. In this paper, we propose a novel Cooperative Caching algorithm for vehicular networks with multi-level federated Reinforcement Learning (named CoCaRL) to dynamically determine which contents should be replaced and where the content requests should be served. In CoCaRL, Deep Reinforcement Learning (DRL) is employed to optimize the cooperative caching policy between RoadSide Units (RSUs) of vehicular networks, while a federated learning framework applies to reduce the computation and communication overhead in a decentralized way. To speed-up the convergence rate, we also develop a two-level aggregation mechanism for federated learning, where the low-level aggregation is performed at the RSUs and the high-level aggregation is executed at a Global Aggregator (GA). Through extensive simulation experiments, we demonstrate that our algorithm can: 1) achieve a higher hit rate than four baseline algorithms, 2) converge faster than original federated reinforcement learning without multi-level aggregation, and 3) perform good adaptability to different cache capacities and content quantities.

Information-Centric Networking is (ICN) gaining increasingly concerns, as an important direction of the future Internet architecture research. To study the impacts of various cache policies on overall performance of ICN network, we formulate the in- network caching problem of ICN into Mixed-Integer Linear Programming problem. Furthermore, we infer that frequency-based cache policies like LFU are supposed to perform well by analyzing the properties of optimal cache assignment, which is corroborated by our simulation results. We attempt to explore the impact of the distance between cache and the original content on cache policy performance by posing a novel cache policy named LB (Least Benefit), which takes into account the distance factor besides frequency. Through extensive simulations under various scenarios and configurations, we find that the performance gain brought by LB is limited in comparison to that of LFU, which implies that more sophisticated cache policies involving the distance factor should be considered to improve LFU. Our simulation results also show that, under a reasonable setting of cache size and request pattern, the average hops to get content can be reduced significantly by nearly 50% in comparison to that of the scenario without in- network caching.

Scalability is of paramount importance in the design of reliable multicast transport protocols, and requires careful consideration of a number of problems such as feedback implosion, retransmission scoping, distributed loss recovery, and congestion control. In this article, we present a reliable multicast architecture that invokes active services at strategic locations inside the network to comprehensively address these challenges. Active services provide the ability to quickly and efficiently recover from loss at the point of loss. They also exploit the physical hierarchy for feedback aggregation and effective retransmission scoping with minimal router support. We present two protocols, one for packet loss recovery and another for congestion control, and describe an experimental testbed where these have been implemented. Analytical and experimental results are used to demonstrate that the active services architecture improves resource usage, reduces latency for loss recovery, and provides TCP-friendly congestion control.

Named Data Networking (NDN) has a built-in caching capability that is enabled with the help of its Content Store. Caching in NDN has several benefits, such as reducing overhead on the producer side, avoiding a single point of failure, and reducing network load. The primary caching policy of the NDN architecture is to leave copies everywhere. However, this scheme induces significant cache redundancy. Existing advanced cache techniques either periodically share the entire list of cached content at a node or make a caching decision without knowing the cached content at other nodes in the network. We propose an intelligent cache policy that estimates cache distribution and topology through a collaborative network of NDN routers, guided by Interest and Data packets. Additionally, we develop an NDN simulator to facilitate the testing and evaluation of our machine learning caching strategy. This paper details the cache policy, the design and functionality of our NDN simulator, and the experiments comparing our approach to baseline caching strategies.

This paper introduces the caching strategy and service policy optimization problem in a cache-satellite distribution service and formulates it as a mathematical optimization problem. A number of system performances like hit rate, cache disk space, bandwidth gain, bandwidth requirements for the satellite channel and latency experienced by the clients were modeled into a mathematical formulation. To show how the optimal caching and service policy responds to the system parameter changes, computational experiments were performed on a hypothetical market having forty ISPs and one billion Web documents. It was found that the correlation between document size and popularity is most influential and the concentration of popularity appeared to be the main factor determining the sensitivity of the optimal caching levels and the corresponding system performances like hit rate and byte hit rate. The achieved hit rate ranged between 64 and 74% with the corresponding caching levels of 25 to 40%. Based on these findings, the importance of caching strategy and service policy optimization was discussed and a number of further research issues were suggested.

A promising solution in order to cope with the massive request of wireless data traffic consists of having replicas of the potential requested content memorized across the network. In cache-enabled heterogeneous networks, content is pre-fetched close to the users during network off-peak periods in order to directly serve the users when the network is congested. In fact, the main idea behind caching is the replacement of backhaul capacity with storage capabilities, for example, at the edge of the network. Caching content at the edge of heterogeneous networks not only leads to significantly reduce the traffic congestion in the backhaul link but also leads to achieve higher levels of energy efficiency. However, the good performance of a system foresees a deep analysis of the possible caching techniques. Due to the physical limitation of the caches’ size and the excessive amount of content, the design of caching policies which define how the content has to be cached and select the likely data to store is crucial. Within this thesis, caching techniques for storing and delivering the content in heterogeneous networks are investigated from two different aspects. The first part of the thesis is focused on the reduction of the power consumption when the cached content is delivered over an Gaussian interference channel and per-file rate constraints are imposed. Cooperative approaches between the transmitters in order to mitigate the interference experienced by the users are analyzed. Based on such approaches, the caching optimization problem for obtaining the best cache allocation solution (in the sense of minimizing the average power consumption) is proposed. The second part of the thesis is focused on caching techniques at packet level with the aim of reducing the transmissions from the core of an heterogeneous network. The design of caching schemes based on rate-less codes for storing and delivering the cached content are proposed. For each design, the placement optimization problem which minimizes the transmission over the backhaul link is formulated.

In this paper, we investigate joint optimization of secure AirComp and reliable multicasting assisted by a multiple-input multiple-output untrusted two-way relay, where artificial noise is employed at the access point (AP) to interfere the relay for ensuring secure AirComp. We aim to minimize the computation distortion at the AP by jointly designing the transmit variables of all the nodes and the aggregation variables at the AP and relay, under the secure AirComp constraint, the reliable multicasting constraint, and the transmit power constraints of all the nodes. We consider two scenarios that perfect and imperfect channel state information (CSI) are available. For the former, the formulated optimization problem is highly nonconvex, and we propose an effective block coordinate descent (BCD)-penalty successive convex approximation (penaltySCA) method to solve the nonconvex problem. For the latter, we model the imperfect CSI by using worst-case criterion and the formulated robust optimization problem is much more challenging than its counterpart with perfect CSI. To solve the robust problem effectively, we first transform it to a deterministic optimization problem by employing some powerful mathematical lemmas, and then apply the proposed BCD-penaltySCA method to solve the reformulated deterministic problem. The proposed methods are shown by simulations to significantly reduce the computation distortion compared with other benchmarks under considering secure AirComp and reliable multicasting.

We propose an in-network caching architecture using Open-Flow to coordinate caching decisions in the network. Our scheme, called CacheFlow, extends the cache-and-forward concept by moving contents closer to the clients hop-by-hop using TCP for sending requests and retrieving contents. As such, CacheFlow can be incrementally implemented and deployed in the real network. In this paper, we present a simulation study of several caching policies, including a random cache policy, a statically optimal cache placement policy and a new disk placement strategy that places popular contents at the center of the network. Experimental results show that simple in-network caching policies can be realized using today's technology to improve network performance.

Caching at base station (BS) has attracted significant research efforts for future wireless networks. Most existing works are based on the assumption of static content catalogue and stationary popularity distribution, which however is far away from the reality as recently reported in the literature. In this paper, we take the popularity dynamics into account, and study the caching policy at BS for a traffic model consisting of two categories of contents respectively with long and short lifespans. By modeling the two categories of contents with Independent Reference Model (IRM) and Shot Noise Model (SNM), we formulate a cache resource allocation problem to maximize the total cache hit ratio for both categories of contents, which gives rise to a hybrid proactive and reactive caching policy. We solve the problem numerically for general case and provide closed-form solutions for several special cases. Numerical and simulation results demonstrate remarkable performance gain of the proposed caching policy over non-hybrid caching polices.