Cache Capacity Research Articles

GraNDe: Efficient Near-Data Processing Architecture for Graph Neural Networks

Graph Neural Network (GNN) models have attracted attention, given their high accuracy in interpreting graph data. One of the primary building blocks of a GNN model is aggregation, which gathers and averages the feature vectors corresponding to the nodes adjacent to each node. Aggregation works by multiplying the adjacency and feature matrices. The size of both matrices exceeds the on-chip cache capacity for many realistic datasets, and the adjacency matrix is highly sparse. These characteristics lead to little data reuse, causing intensive main-memory accesses during the aggregation process. Thus, aggregation exhibits memory-intensive characteristics and dominates most of the total execution time. In this paper, we propose GraNDe, an NDP architecture that accelerates memory-intensive aggregation operations by locating NDP modules near DRAM datapath to exploit rank-level parallelism. GraNDe maximizes bandwidth utilization by separating the memory channel path with the buffer chip in between so that pre-/post-processing in the host processor and reduction in NDP modules operate simultaneously. By exploring the preferred data mappings of the operand matrices to DRAM ranks, we architect GraNDe to support adaptive matrix mapping that applies the optimal mapping for each layer depending on the dimension of the layer and the configuration of a memory system. We also propose adj-bundle broadcasting and re-tiling optimizations to reduce the transfer time for adjacency matrix data and to improve feature vector data reusability by exploiting tiling with consideration of adjacency between nodes. GraNDe achieves 3.01× and 1.69× on average, and up to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$4.00\times$</tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$1.98\times$</tex-math></inline-formula> speedups of GCN aggregation over the baseline system and the state-of-the-art NDP architecture for GCN, respectively.

Joint optimization for service-caching, computation-offloading, and UAVs flight trajectories over rechargeable UAV-aided MEC using hierarchical multi-agent deep reinforcement learning

Due to the high mobility, high chance of line-of-sight (LoS) transmission, and flexible deployment, unmanned aerial vehicles (UAVs) have been used as mobile edge computing (MEC) servers to provide ubiquitous computation services to mobile users (MUs). However, the limited energy storage, caching capacity, and computation resources of UAVs bring new challenges for UAV-aided MEC, e.g., how to recharge UAVs and how to jointly optimize service-caching, computation-offloading, and UAVs flight trajectories. To overcome the above-mentioned difficulties, in this paper we study the joint optimization for service-caching, computation-offloading, and UAVs flight trajectories for UAV-aided MEC, where multiple rechargeable UAVs cooperatively provide MEC services to a number of MUs. First, we formulate an energy minimization problem to minimize all MUs' energy consumptions by taking into account the mobility of MUs and the energy replenishment of UAVs. Then, using the hierarchical multi-agent deep reinforcement learning (HMDRL), we develop a two-timescale based joint service-caching, computation-offloading, and UAVs flight trajectories scheme, called HMDRL-Based SCOFT. Using HMDRL-Based SCOFT, we derive UAVs' service-caching policies in each time frame, and then derive UAVs flight trajectories and MUs' computation-offloading in each time slot. Finally, we validate and evaluate the performances of our proposed HMDRL-Based SCOFT scheme through extensive simulations, which show that our developed scheme outperforms the other baseline schemes to converge faster and greatly reduce MUs' energy consumptions.

Research on paging enhancements for 5G-A downlink transmission energy saving

5G-Advanced (5G-A), an evolutionary iteration of 5G, effectively enhances 5G services. The increasing complexity in downlink services scenarios stresses the necessity for research into the integration of efficient communication with low-carbon solutions. Historically, there has been an emphasis on reliability and precision, at the expense of power consumption. Although energy-saving technologies like Idle mode-Discontinuous Reception (IDRX) and Paging Early Indication (PEI) have been introduced to reduce power consumption in UE, they have not been fully tailored to the paging characteristics of 5G-A downlink services. In this paper, we take full account of the impact of paging message density on energy saving measures and propose an enhanced paging technology, termed Predictive-PEI (PPEI), which is designed to reduce UE overhead while minimizing latency whenever possible. Towards this end, we design a dual threshold decision framework founded on machine learning, mainly involving two steps. We first use the LSTM-FNN neural network to forecast the arrival moment of upcoming paging messages based on past real information. Then, the output of the initial prediction is as the input of the next dual threshold decision algorithm, to determine the optimal moment for transmitting the PEI. The restrictive factors, encompass average delay threshold and cache capacity threshold, playing a role in decisions regarding paging message caching and decoding. Compared to the existing schemes, our PPEI scheme flexibly sends efficient PEI according to the actual paging characteristics by introducing machine learning, resulting in substantial power savings of up to 38.89% while concurrently ensuring effective latency control.

Optimizing storage on fog computing edge servers: A recent algorithm design with minimal interference.

The burgeoning field of fog computing introduces a transformative computing paradigm with extensive applications across diverse sectors. At the heart of this paradigm lies the pivotal role of edge servers, which are entrusted with critical computing and storage functions. The optimization of these servers' storage capacities emerges as a crucial factor in augmenting the efficacy of fog computing infrastructures. This paper presents a novel storage optimization algorithm, dubbed LIRU (Low Interference Recently Used), which synthesizes the strengths of the LIRS (Low Interference Recency Set) and LRU (Least Recently Used) replacement algorithms. Set against the backdrop of constrained storage resources, this research endeavours to formulate an algorithm that optimizes storage space utilization, elevates data access efficiency, and diminishes access latencies. The investigation initiates a comprehensive analysis of the storage resources available on edge servers, pinpointing the essential considerations for optimization algorithms: storage resource utilization and data access frequency. The study then constructs an optimization model that harmonizes data frequency with cache capacity, employing optimization theory to discern the optimal solution for storage maximization. Subsequent experimental validations of the LIRU algorithm underscore its superiority over conventional replacement algorithms, showcasing significant improvements in storage utilization, data access efficiency, and reduced access delays. Notably, the LIRU algorithm registers a 5% increment in one-hop hit ratio relative to the LFU algorithm, a 66% enhancement over the LRU algorithm, and a 14% elevation in system hit ratio against the LRU algorithm. Moreover, it curtails the average system response time by 2.4% and 16.5% compared to the LRU and LFU algorithms, respectively, particularly in scenarios involving large cache sizes. This research not only sheds light on the intricacies of edge server storage optimization but also significantly propels the performance and efficiency of the broader fog computing ecosystem. Through these insights, the study contributes a valuable framework for enhancing data management strategies within fog computing architectures, marking a noteworthy advancement in the field.

Performance Analysis of Internet of Vehicles Mesh Networks Based on Actual Switch Models

The rapid growth of the automotive industry has exacerbated the conflict between the complex traffic environment, increasing communication demands, and limited resources. Given the imperative to mitigate traffic and network congestion, analyzing the performance of Internet of Vehicles (IoV) mesh networks is of great practical significance. Most studies focus solely on individual performance metrics and influencing factors, and the adopted simulation tools, such as OPNET, cannot achieve the dynamic link generation of IoV mesh networks. To address these problems, a network performance analysis model based on actual switches is proposed. First, a typical IoV mesh network architecture is constructed and abstracted into a mathematical model that describes how the link and topology changes over time. Then, the task generation model and the task forwarding model based on actual switches are proposed to obtain the real traffic distribution of the network. Finally, a scientific network performance indicator system is constructed. Simulation results demonstrate that, with rising task traffic and decreasing node caching capacity, the packet loss rate increases, and the task arrival rate decreases in the network. The proposed model can effectively evaluate the network performance across various traffic states and provide valuable insights for network construction and enhancement.

Hardware Compression Method for On-Chip and Interprocessor Networks with Wide Channels and Wormhole Flow Control Policy

Increasing the number of processing cores is currently a common way to boost processor performance. However, the load on the memory subsystem consequently increases as the number of its agents grows. Hardware data compression is an unconventional approach to improving memory subsystem performance by reducing, firstly, the main memory access rate by increasing the cache capacity and, secondly, data traffic by packing the data more densely. The paper describes the implementation of hardware data compression in the on-chip network and interprocessor links of a configuration with wide data transmission channels and a wormhole flow control policy. The existing solutions cannot be applied to such configurations because they are essentially based on using narrow data channels and flow control policies implying uninterrupted packet transmission, which is not maintained with the wormhole flow control. The method proposed in this paper enables the use of hardware compression in the aforementioned configuration by moving data compression and decompression from networks to the connected devices, as well as by using a number of optimizations to hide the data processing delays. Optimizations of some specific cases, such as the transmission of large data packets with several cache lines or the transmission of zero data, are considered. Special attention is given to data transmission via interprocessor links, where, due to their lower bandwidth compared to the on-chip network, data compression can be the most beneficial. The increase in memory subsystem bandwidth from using hardware data compression was confirmed in the experiments showing the relative IPC increase in SPEC CPU2017 benchmarks up to 14 percent.

Multilayered Decentralized Coded Caching With Nonuniform Popularity and Multilevel Cache Capacity in Space–Air–Ground Integrated Networks

Multilayered Decentralized Coded Caching With Nonuniform Popularity and Multilevel Cache Capacity in Space–Air–Ground Integrated Networks

Application-Level Caching Approach Based on Enhanced Aging Factor and Pearson Correlation Coefficient

Relational database management systems (RDBMS) have long served as the fundamental infrastructure for web applications. Relatively slow access speeds characterize an RDBMS because its data is stored on a disk. This RDBMS weakness can be overcome using an in-memory database (IMDB). Each query result can be stored in the IMDB to accelerate future access. However, due to the limited capacity of the server cache in the IMDB, an appropriate data priority assessment mechanism needs to be developed. This paper presents a similar cache framework that considers four data vectors, namely the data size, timestamp, aging factor, and controller access statistics for each web page, which serve as the foundation elements for determining the replacement policy whenever there is a change in the content of the server cache. The proposed similarCache employs the Pearson correlation coefficient to quantify the similarity levels among the cached data in the server cache. The lowest Pearson correlation coefficients cached data are the first to be evicted from the memory. The proposed similarCache was empirically evaluated based on simulations conducted on four IRcache datasets. The simulation outcomes revealed that the data access patterns, and the configuration of the allocated memory cache significantly influenced the hit ratio performance. In particular, the simulations on the SV dataset with the most minor memory space configuration exhibited a 2.33% and 1% superiority over the SIZE and FIFO algorithms, respectively. Future tasks include building a cache that can adapt to data access patterns by determining the standard deviation. The proposed similarCache should raise the Pearson coefficient for often available data to the same level as most accessed data in exceptional cases.

Optimized Two-Tier Caching with Hybrid Millimeter-Wave and Microwave Communications for 6G Networks

Data caching is a promising technique to alleviate the data traffic burden from the backhaul and minimize data access delay. However, the cache capacity constraint poses a significant challenge to obtaining content through the cache resource that degrades the caching performance. In this paper, we propose a novel two-tier caching mechanism for data caching on mobile user equipment (UE) and the small base station (SBS) level in ultra-dense 6G heterogeneous networks for reducing data access failure via cache resources. The two-tier caching enables users to retrieve their desired content from cache resources through device-to-device (D2D) communications from neighboring users or the serving SBS. The cache-enabled UE exploits millimeter-wave (mmWave)-based D2D communications, utilizing line-of-sight (LoS) links for high-speed data transmission to content-demanding mobile UE within a limited connection time. In the event of D2D communication failures, a dual-mode hybrid system, combining mmWave and microwave μWave technologies, is utilized to ensure effective data transmission between the SBS and UE to fulfill users’ data demands. In the proposed framework. the data transmission speed is optimized through mmWave signals in line-of-sight (LoS) conditions. In non-LoS scenarios, the system switches to μWave mode for obstacle-penetrating signal transmission. Subsequently, we propose a reinforcement learning (RL) approach to optimize cache decisions through the approximation of the Q action-value function. The proposed technique undergoes iterative learning, adapting to dynamic network conditions to enhance the content placement policy and minimize delay. Extensive simulations demonstrate the efficiency of our proposed approach in significantly reducing network delay compared with benchmark schemes.

Enhancing Heterogeneous Network Performance: Advanced Content Popularity Prediction and Efficient Caching

With the popularity of smart devices and the growth of high-bandwidth applications, the wireless industry is facing an increased surge in data traffic. This challenge highlights the limitations of traditional edge-caching solutions, especially in terms of content-caching effectiveness and network-communication latency. To address this problem, we investigated efficient caching strategies in heterogeneous network environments. The caching decision process becomes more complex due to the heterogeneity of the network environment, as well as due to the diversity of user behaviors and content requests. To address the problem of increased system latency due to the dynamically changing nature of content popularity and limited cache capacity, we propose a novel content placement strategy, the long-short-term-memory–content-population-prediction model, to capture the correlation of request patterns between different contents and the periodicity in the time domain, in order to improve the accuracy of the prediction of content popularity. Then, to address the heterogeneity of heterogeneous network environments, we propose an efficient content delivery strategy: the multi-intelligent critical collaborative caching policy. This strategy models the edge-caching problem in heterogeneous scenarios as a Markov decision process using multi-base-station-environment information. In order to fully utilize the multi-intelligence information, we have improved the actor–critic approach by integrating the attention mechanism into a neural network. Whereas the actor network is responsible for making decisions based on local information, the critic network evaluates and enhances the actor’s performance. We conducted extensive simulations, and the results showed that the Long Short Term Memory content population prediction model was more advantageous, in terms of content-popularity-prediction accuracy, with a 28.61% improvement in prediction error, compared to several other existing methods. The proposed multi-intelligence actor–critic collaborative caching policy algorithm improved the cache-hit-rate metric by up to 32.3% and reduced the system latency by 1.6%, demonstrating the feasibility and effectiveness of the algorithm.

An Efficient Block Successive Upper-Bound Minimization Algorithm for Caching a Reconfigurable Intelligent Surface-Assisted Downlink Non-Orthogonal Multiple Access System

With the booming rollout of 5G communication, abundant new technologies have been proposed for quality of service requirements. In terms of the betterment in transmission coverage, mobile edge caching (MEC) has shown potential in reducing the transmission outage. The performance of MEC, meanwhile, can be promisingly enhanced by reconfigurable intelligent surfaces (RIS). Under this context, we explore a system comprising a small base-station (SBS) with limited cache capacity, two users, and one RIS. The SBS transmits the contents from the cache or fetches them from the remote backhaul hub to communicate with users through directional and possibly reflective channels. In this point-to-multipoint connection, non-orthogonal multiple access (NOMA) is applied, improving the capacity of the system. To minimize the outage probability, we first propose a caching policy from entropy perspective, based on which we investigate the beamforming and power allocation problem. The issue, however, is non-convex and involves multi-dimensional optimization. To address this, we introduce an efficient block successive upper-bound minimization algorithm, grounded in Gershgorin’s circle theorem. This algorithm aims to find the globally optimal solution for power allocation and RIS beamformer, considering both the channel condition and content popularity. Numerical studies are performed to verify the effectiveness of the proposed algorithm.

Efficient content caching for 5G assisted vehicular networks

Traffic congestion is one of the major challenges faced by daily commuters in smart cities. An autonomous transportation system with a 5 G-based Cellular Vehicle-to-Everything (C-V2X) communication system is the solution to meet the traffic challenges faced in smart cities. Vehicular networks provide wireless connectivity to enable a large number of connected vehicle applications. Vehicular networks allow vehicles to share their emergency and infotainment traffic by following vehicle-to-vehicle (V2V) or by using vehicle-to-infrastructure (V2I) communication. The infrastructure of vehicular networks mainly comprises multiple Road Side Units (RSUs). Fog computing nodes are placed adjacent to these RSUs to provide quick access to vehicles. For infotainment traffic, vehicles intend to download their required content from the content provider. Caching the same contents from the nearby fog computing node significantly reduces delay with improved quality of service. As there are millions of contents with varying sizes, caching all demanded contents on these fog nodes is not possible due to their limited caching capacity. In this work, we propose an improved content caching scheme for fog nodes to satisfy vehicles and content providers for fair content placement. The proposed algorithm is based on a modified Gale–Shapley technique that considers factors such as content popularity, vehicle connectivity, and quality of the communication channel to optimize the content caching process. Simulation results show that the proposed technique caches a higher number of popular contents with lower downloading time.

Controller-driven vector autoregression model for predicting content popularity in programmable named data networking devices.

Named Data Networking (NDN) has emerged as a promising network architecture for content delivery in edge infrastructures, primarily due to its name-based routing and integrated in-network caching. Despite these advantages, sub-optimal performance often results from the decentralized decision-making processes of caching devices. This article introduces a paradigm shift by implementing a Software Defined Networking (SDN) controller to optimize the placement of highly popular content in NDN nodes. The optimization process considers critical networking factors, including network congestion, security, topology modification, and flowrules alterations, which are essential for shaping content caching strategies. The article presents a novel content caching framework, Popularity-aware Caching in Popular Programmable NDN nodes (PaCPn). Employing a multi-variant vector autoregression (VAR) model driven by an SDN controller, PaCPn periodically updates content popularity based on time-series data, including 'request rates' and 'past popularity'. It also introduces a controller-driven heuristic algorithm that evaluates the proximity of caching points to consumers, considering factors such as 'distance cost,' 'delivery time,' and the specific 'status of the requested content'. PaCPn utilizes customized DATA named packets to ensure the source stores content with a valid residual freshness period while preventing intermediate nodes from caching it. The experimental results demonstrate significant improvements achieved by the proposed technique PaCPn compared to existing schemes. Specifically, the technique enhances cache hit rates by 20% across various metrics, including cache size, Zipf parameter, and exchanged traffic within edge infrastructure. Moreover, it reduces content retrieval delays by 28%, considering metrics such as cache capacity, the number of consumers, and network throughput. This research advances NDN content caching and offers potential optimizations for edge infrastructures.

Joint Trajectory Design and Resource Optimization in UAV-Assisted Caching-Enabled Networks with Finite Blocklength Transmissions

In this study, we design and analyze a reliability-oriented downlink wireless network assisted by unmanned aerial vehicles (UAVs). This network employs non-orthogonal multiple access (NOMA) transmission and finite blocklength (FBL) codes. In the network, ground user equipments (GUEs) request content from a remote base station (BS), and there are no direct connections between the BS and the GUEs. To address this, we employ a UAV with a limited caching capacity to assist the BS in completing the communication. The UAV can either request uncached content from the BS and then serve the GUEs or directly transmit cached content to the GUEs. In this paper, we first introduce the decoding error rate within the FBL regime and explore caching policies for the UAV. Subsequently, we formulate an optimization problem aimed at minimizing the average maximum end-to-end decoding error rate across all GUEs while considering the coding length and maximum UAV transmission power constraints. We propose a two-step alternating optimization scheme embedded within a deep deterministic policy gradient (DDPG) algorithm to jointly determine the UAV trajectory and transmission power allocations, as well as blocklength of downloading phase, and our numerical results show that the combined learning-optimization algorithm efficiently addresses the considered problem. In particular, it is shown that a well-designed UAV trajectory, relaxing the FBL constraint, increasing the cache size, and providing a higher UAV transmission power budget all lead to improved performance.

Research on high performance web service construction method based on JavaScript asynchronous programming technique

Abstract In the face of massive concurrent user access in the era of big data, how to build high-performance web services has become one of the difficulties to be solved by network applications. This paper utilizes Node.js architecture to fully utilize the event-based programming and non-blocking I/O characteristics of JavaScript language to improve the CPU utilization of the web service side. It also proposes the strategy of predicting user-accessed resources for cache replacement through a Gaussian mixture model to further optimize the performance of high-concurrency web services. The performance of the high concurrency web framework constructed using Node.js technology plays excellent, the performance of multi-threaded Node.js is better than single-threaded, and compared with Apache architecture, the average response time, request-response rate, and data throughput test of Node.js architecture in high concurrency scenarios is more advantageous than Apache architecture. The Gaussian mixture clustering model is effective in dealing with data with high-dimensional features, and the average accuracy of the four clustering processes of each algorithm is, in descending order, GMM (79.67%), BIRCH (78.09%), and K-means (77.52%). In addition, the cache replacement strategy based on Gaussian Mixture Model is more effective, with an accuracy rate of nearly 80% and a byte hit rate of nearly 45% when the cache capacity reaches 214KB, which are both higher than the traditional cache replacement strategy with the same cache capacity.

Resource Allocation in UAV-Enabled NOMA Networks for Enhanced Six-G Communications Systems

Enhancing energy efficiency, content distribution, latency, and transmission speeds are vital components of communication systems. Multiple access methods hold great promise for boosting these performance indicators. This manuscript evaluates the effectiveness of Non-Orthogonal Multiple Access (NOMA) and Orthogonal Multiple Access (OMA) systems within a single cell, where users are scattered randomly and rely on relays for dependability. This paper presents a model for improving energy efficiency, content distribution, latency, and transmission speeds in communication systems using NOMA and OMA systems within a single cell. Additionally, this paper also proposes a caching strategy using unmanned aerial vehicles (UAVs) as aerial base stations for ground users. These UAVs distribute cached content to minimize the overall latency of content demands from ground users while modifying their positions. We carried out simulations using various cache capacities and user counts linked to their respective UAVs. Furthermore, we evaluated OMA and NOMA in terms of the achievable rate and energy efficiency. The proposed model has achieved noteworthy enhancement across various scenarios including different sum rates, numbers of mobility users, diverse cache sizes, and amounts of power allocation.

ENCache: Improving content delivery with cooperative caching in Named Data Networking

Content centric networks offer better response time and resilience to failures by caching the content in the network routers. Owing to the limited capacity of caches and the quest for maximizing the utilization of available caches has lead to many interesting caching techniques. While on-path caching is a commonly adopted caching technique to reduce content access latency, it does not guarantee better cache hit ratio when considered across multiple receivers in the network. In this paper, we present eNCache which is a cooperative caching method that cooperates among the routers in the neighborhood to improve overall performance. It retains the simplicity of on-path caching and makes caching, routing (queries and replies) decisions by interacting with off-path routers in the neighborhood. We present the working of eNCache with algorithms for request handling and caching decisions at every router. We also formalize the interaction between a router and its neighbors by providing an extension to the Named Data Networking Interest packet structure. We evaluate and compare the performance of eNCache with other popular on-path caching strategies by performing simulations on standard RocketFuel Internet Service Provider (ISP) topologies using a discrete event simulator. We also compare the performance of a variant of enhanced NCache, i.e.,eNCache with hash based off-path caching methods. Our simulation results show that eNCache has reduced latency, better cache hit ratio and diversity.

Analysis and Optimization of Cache-Enabled mmWave HetNets With Integrated Access and Backhaul

In millimeter wave heterogeneous networks with integrated access and backhaul (mABHetNets), a considerable part of spectrum resources are occupied by the backhaul link, which limits the performance of the access link. In order to overcome such backhaul “ <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">spectrum occupancy</i> ”, we introduce cache into mABHetNets. Caching popular files at small base stations (SBSs) can offload the backhaul traffic and transfer spectrum from the backhaul link to the access link. To approach the optimal performance of cache-enabled mABHetNets, we first analyze the signal-to-interference-plus-noise ratio distribution and derive the average potential throughput (APT) expression by stochastic geometric tools. Then, based on our analytical work, we formulate a joint optimization problem of cache decision and spectrum partition to maximize the APT. Inspired by the block coordinate descent (BCD) method, we propose a joint cache decision, spectrum partition and power allocation (JCSPA) algorithm to approach the optimal solution. Numerical results show the effectiveness and enhancement of the proposed algorithm. Besides, we verify the APT under different parameters and find that the introduction of cache facilitates the transfer of backhaul spectrum to access link. Jointly deploying appropriate caching capacity at SBSs and performing specified spectrum partition can bring up about 90% APT gain in mABHetNets.

Catalyst: Optimizing Cache Management for Large In-memory Key-value Systems

In-memory key-value cache systems, such as Memcached and Redis, are essential in today's data centers. A key mission of such cache systems is to identify the most valuable data for caching. To achieve this, the current system design keeps track of each key-value item's access and attempts to make accurate estimation on its temporal locality. All it aims is to achieve the highest cache hit ratio. However, as cache capacity quickly increases, the overhead of managing metadata for a massive amount of small key-value items rises to an unbearable level. Put it simply, the current fine-grained, heavy-cost approach cannot continue to scale. In this paper, we have performed an experimental study on the scalability challenge of the current key-value cache system design and quantitatively analyzed the inherent issues related to the metadata operations for cache management. We further propose a key-value cache management scheme, called Catalyst , based on a highly efficient metadata structure, which allows us to make effective caching decisions in a scalable way. By offloading non-essential metadata operations to GPU, we can further dedicate the limited CPU and memory resources to the main service operations for improved throughput and latency. We have developed a prototype based on Memcached. Our experimental results show that our scheme can significantly enhance the scalability and improve the cache system performance by a factor of up to 4.3.

OA-Cache: Oracle Approximation-Based Cache Replacement at the Network Edge

With the explosive increase in mobile data traffic generated by various application services like video-on-demand and stringent quality of experience requirements of users, mobile edge caching is a promising paradigm to reduce delivery latency and network congestions by serving content requests locally. However, how to conduct cache replacement when the cache is full is a challenging issue when faced with enormous content volume and limited cache capacity at the network edge while the future request pattern is unknown ahead. In this paper, we propose a cache replacement algorithm based on the oracle approximation named OA-Cache in an end-to-end manner to maximize the cache hit rate. Specifically, we construct a complex model that uses a temporal convolutional network to capture the long and short dependencies between content requests. Then, an attention mechanism is adopted to find out the correlations between requests in the sliding window and cached contents. Instead of training a policy to mimic Belady that evicts the content with the longest reuse distance, we cast the learning task into a classification model to distinguish unpopular contents from popular ones. Finally, we apply the knowledge distillation approach to assist in transferring knowledge from a large pre-trained complex network to a lightweight network to readily accommodate to the network edge scenario. To validate the effectiveness of OA-Cache, we conduct extensive experiments on real-world datasets. The evaluation results demonstrate that OA-Cache can achieve better performance compared to candidate algorithms.