Caching System Research Articles

Moving the California distributed CMS XCache from bare metal into containers using Kubernetes

The University of California system maintains excellent networking between its campuses and a number of other Universities in California, including Caltech, most of them being connected at 100 Gbps. UCSD and Caltech Tier2 centers have joined their disk systems into a single logical caching system, with worker nodes from both sites accessing data from disks at either site. This successful setup has been in place for the last two years. However, coherently managing nodes at multiple physical locations is not trivial and requires an update on the operations model used. The Pacific Research Platform (PRP) provides Kubernetes resource pool spanning resources in the science demilitarized zones (DMZs) in several campuses in California and worldwide. We show how we migrated the XCache services from bare-metal deployments into containers using the PRP cluster. This paper presents the reasoning behind our hardware decisions and the experience in migrating to and operating in a mixed environment.

An Incremental Learning Based Edge Caching System: From Modeling to Evaluation

Caches are widely applied to improve data delivery performance in distributed systems like edge networks and content delivery networks (CDNs). We consider caching mechanism in those networks that deliver contents to end users. The challenge comes from the dynamic content distribution problem. The distribution of data popularity is highly skewed and changing over time. Besides, the access pattern of the user requests also varies over time. Some learning algorithms for edge caching problems need to rebuild a new model periodically to adapt to system dynamics, where the knowledge learned from the past is discarded. Besides, each model updating needs a large amount of data, leading to outdated models for consecutive user requests. Inspired by the success of incremental learning approaches in processing massive data in real time, we propose an incremental learning based framework at an edge caching server. The incremental learning algorithm is used to preserve valuable knowledge and to adapt to dynamic workloads faster. We implement our incremental learning based cache system prototype and evaluate its performance under various real-world workloads. The experimental results show that our algorithm can boost cache hit ratio for dynamic workloads compared with the state-of-the-art caching algorithms.

A Stepwise Rate-Compatible LDPC and Parity Management in NAND Flash Memory-Based Storage Devices

The storage capacity of the NAND flash memory has increased rapidly, and accordingly, the error rate for data writing and reading to the flash memory cell has also escalated. Error-correcting code (ECC) modules, such as low-density parity-check (LDPC), have been applied to flash controllers for error recovery. However, since the error rate increases rapidly, compared to the aging factor and program/erase (P/E) cycle, fixed ECCs and parities are inappropriate methods for resolving this proliferating error, according to the P/E cycle. Therefore, the design of a dynamic ECC scheme and a proper ECC parity management system to increase the lifespan of flash memory storage devices remains in great demand. Herein, an LDPC encoding and decoding scheme is designed to obtain a step-by-step code rate according to the P/E cycle by applying a stepwise rate-compatible LDPC. In addition, an ECC parity management scheme for the increasingly excessive stage-wise ECC is proposed to reduce management and read/write operational overheads. The ECC management scheme also includes the ECC cache system. The proposed LDPC, as well as its management system, will improve the recovery ability of the NAND flash storage device according to the P/E cycle, while it can reduce system read and write overheads due to additional parity data growth.

Peak-Hour Caching Schemes of Mobile Devices for Overload Cells in Wireless Caching Systems

In the case of broadband wireless communication systems with high carrier frequencies, high-speed wireless communication is feasible, but if high-capacity backhaul is not supplied to all base stations, a backhaul bottleneck may occur. To solve this backhaul bottleneck problem, one can use wireless caching or device-to-device (D2D) caching technology that installs a cache on a mobile end device, use the cache to store video content, and then supply the video content when requested by other end devices. Prior to peak traffic hours, video content can be stored on mobile end devices according to the content preferences, but it is difficult to solve cell overload problems that occur during peak traffic hours. This is because peak traffic hours can persist for a long time, and thus device positions, content preferences, and cell loads are liable to change during peak traffic hours. Providing broadband radio and high-capacity backhaul to all cells can create cost issues, but if there are some cells with broadband wireless resources and high-capacity backhaul, these cells may often support sufficient data even during peak traffic hours. If a mobile end device can intermittently enter a cell with sufficient data supply, it can update its cache using some short-term information. In this paper, we compare wireless caching schemes that update the caches of mobile end devices during peak traffic hours with the goal of reducing overload data. If high-capacity base stations are installed within a certain distance interval from each other to reduce the content update period, then good performance can be obtained by frequently updating the caches while considering the short-term content preferences of nearby overload cells. In addition, if a mobile device can predict its movement path, performance can be further improved by considering only the overload cells in the predicted path.

Optimizing Content Placement and Delivery in Wireless Distributed Cache Systems Through Belief Propagation

Wireless distributed cache system (WDCS) is a promising technique to offload traffic and improve energy efficiency in content-centric networks. In this paper, the content sharing problem is considered by minimizing the average energy consumption for content caching and delivering in device-to-device (D2D) enabled cellular networks. The problem is formulated as a joint optimization problem of multiple interactive variables, which is NP-hard. In order to solve this problem, we first decompose the original problem into three subproblems, which are content helper (CH) selection, content caching, and content delivering. Specifically, in CH selection, we propose a PageRank based distributed algorithm to estimate the centrality of mobile users (MUs) by jointly analyzing their social relationships and channel conditions. Based on the selected CHs, the content caching and content delivering are formulated into two weighted matching problems. Two belief propagation (BP) based distributed algorithms are then proposed by considering content popularity and channel states to solve the aforementioned problems. Furthermore, based on the correlation of the three subproblems, a heuristic-based alternating iterative optimization method (HAIOM) is proposed to improve the average energy consumption performance of the content delivering of the WDCS. Simulation results show that by jointly optimizing the CH selection, content caching, and content delivering, the proposed scheme is capable of reducing the average energy consumption and the average content delivery delay, as well as of increasing the caching hit ratio.

Content Prefetching of Mobile Caching Devices in Cooperative D2D Communication Systems

The increase of mobile traffic volume in a wireless network using high carrier frequency can lead to coverage and backhaul bottleneck issues, and these issues can be solved by the use of device-to-device (D2D) caching technology that installs a content cache on a device and uses D2D communication to supply contents to other devices. In this paper, a special device that can supply contents using D2D communication to other devices is called a helper, while a device that can receive contents from a helper is called user equipment (UE). Even though a helper may have a large storage capacity compared to a UE, the cache capacity of a helper is still limited, making it difficult to obtain a sufficient hit ratio using a caching scheme based on the long-term content popularity of a large number of UEs while considering the distant future. In this paper, to solve the problem of the conventional D2D caching system, we suggest instead updating the cache contents based on the short-term content preference of individual user for the near future. In the proposed caching scheme, a UE in an area with insufficient data supply predicts the content that will be consumed in the near future. This information is then passed to a mobile helper moving along a predictable travel path, such as a bus and a self-driving car that is passing through or will pass though an area with sufficient data supply. Then, the mobile helper prefetches the contents and moves to update the cache of the nomadic or fixed helper near the UE. The simulation results show that content prefetching considering each UE and the near future can achieve better performance than content caching considering a large number of UEs and the distant future.

Load Imbalance and Caching Performance of Sharded Systems

Sharding is a method for allocating data items to nodes of a distributed caching or storage system based on the result of a hash function computed on the item’s identifier. It is ubiquitously used in key-value stores, CDNs and many other applications. Despite considerable work that has focused on the design and implementation of such systems, there is limited understanding of their performance in realistic operational conditions from a theoretical standpoint. In this paper we fill this gap by providing a thorough modeling of sharded caching systems, focusing particularly on load balancing and caching performance aspects. Our analysis provides important insights that can be applied to optimize the design and configuration of sharded caching systems.

Analysis of an SDN-Based Cooperative Caching Network with Heterogeneous Contents

The ubiquity of data-enabled mobile devices and wireless-enabled data applications has fostered the rapid development of wireless content caching, which is an efficient approach to mitigating cellular traffic pressure. Considering the content characteristics and real caching circumstances, a software-defined network (SDN)-based cooperative caching system is presented. First, we define a new file block library with heterogeneous content attributes [file popularity, mobile user (MU) preference, file size]. An SDN-based three-tier caching network is presented in which the base station supplies control coverage for the entire macrocell and cache helpers (CHs), MUs with cache capacities offer data coverage. Using the ‘most popular content’ and ‘largest diversity content’, a distributed cooperative caching strategy is proposed in which the caches of the MUs store the most popular contents of the file block library to mitigate the effect of MU mobility, and those of the CHs store the remaining contents in a probabilistic caching manner to enrich the content diversity and reduce the MU caching pressure. The request meet probability (RMPro) is subsequently proposed, and the optimal caching distribution of the contents in the probabilistic caching strategy is obtained via optimization. Finally, using the result of RMPro optimization, we also analyze the content retrieval delays that occur when a typical MU requests a file block or a whole file. Simulation results demonstrate that the proposed caching system can achieve quasi-optimal revenue performance compared with other contrasting schemes.

Double Coded Caching in Ultra Dense Networks: Caching and Multicast Scheduling via Deep Reinforcement Learning

Proposed by Maddah-Ali and Niesen, a coded caching scheme has been verified to alleviate the load of networks efficiently. Recently, a new technique called placement delivery array (PDA) was proposed to characterize the coded caching scheme. In this paper, we consider a caching system in the scope of ultra dense networks (UDNs). Each base station (BS) has a finite cache and stores some contents. We propose an efficient coded content caching scheme called double coded caching to make the transmission robust to in-and-out wireless network quality. Then the dynamic caching and multicast scheduling are considered to jointly minimize the average delay and power of the content-centric wireless networks. This stochastic optimization problem can be formulated as a Markov decision process (MDP) with unknown transition probabilities and large state space. We propose a deep reinforcement learning approach to deal with the decision problem. Our algorithm uses a variational auto-encoder (VAE) neural network to approximate the state sufficiently, and uses a weighted double Q-learning scheme to reduce variance and overestimation of the Q function. Numerical results demonstrate that the proposed double coded caching scheme increases the probability of the successful transmission, and the caching and scheduling policy can effectively reduce the delay and the power consumption.

Deep Reinforcement Learning for Adaptive Caching in Hierarchical Content Delivery Networks

Caching is envisioned to play a critical role in next-generation content delivery infrastructure, cellular networks, and Internet architectures. By smartly storing the most popular contents at the storage-enabled network entities during off-peak demand instances, caching can benefit both network infrastructure as well as end users, during on-peak periods. In this context, distributing the limited storage capacity across network entities calls for decentralized caching schemes. Many practical caching systems involve a parent caching node connected to multiple leaf nodes to serve user file requests. To model the two-way interactive influence between caching decisions at the parent and leaf nodes, a reinforcement learning (RL) framework is put forth. To handle the large continuous state space, a scalable deep RL approach is pursued. The novel approach relies on a hyper-deep Q-network to learn the Q-function, and thus the optimal caching policy, in an online fashion. Reinforcing the parent node with ability to learn-and-adapt to unknown policies of leaf nodes as well as spatio-temporal dynamic evolution of file requests, results in remarkable caching performance, as corroborated through numerical tests.

Collaboration Improves the Capacity of Mobile Edge Computing

This article studies the capacity of edge caching systems. Capacity is analyzed from the perspective of node mobility, caching, content popularity, etc., neglecting the influence of edge node cooperation on system performance. However, cooperation among edge nodes has been shown to substantially improve the system performance at the expense of a cooperation cost. Therefore, we attempt to maximize the capacity of mobile edge computing (MEC) subject to a budget constraint on the cooperation cost. We first transform the capacity maximization problem into a transmission distance minimization problem; then, we explore the average transmission distance of each source-to-destination pair under the assumption that the locations of the edge nodes follow a Poisson point process (PPP). We find that both the average transmission distance and the cooperation cost are associated with a key parameter, i.e., the number of content copies. We use the Lagrangian multiplier method to calculate the optimal copy number and propose a file allocation algorithm to store these copies. Finally, we analyze the influence of various parameters on the system capacity. The numerical results verify the efficiency of our solution compared with classic works.

Deep Reinforcement Learning Approaches for Content Caching in Cache-Enabled D2D Networks

Internet of Things (IoT) technology suffers from the challenge that rare wireless network resources are difficult to meet the influx of a huge number of terminal devices. Cache-enabled device-to-device (D2D) communication technology is expected to relieve network pressure with the fact that the requesting contents can be easily obtained from nearby users. However, how to design an effective caching policy becomes very challenging due to the limited content storage capacity and the uncertainty of user mobility pattern. In this article, we study the jointly cache content placement and delivery policy for the cache-enabled D2D networks. Specifically, two potential recurrent neural network approaches [the echo state network (ESN) and the long short-term memory (LSTM) network] are employed to predict users’ mobility and content popularity, so as to determine which content to cache and where to cache. When the local cache of the user cannot satisfy its own request, the user may consider establishing a D2D link with the neighboring user to implement the content delivery. In order to decide which user will be selected to establish the D2D link, we propose the novel schemes based on deep reinforcement learning to implement the dynamic decision making and optimization of the content delivery problems, aiming at improving the quality of experience of overall caching system. The simulation results suggest that the cache hit ratio of the system can be well improved by the proposed content placement strategy, and the proposed content delivery approaches can effectively reduce the request content delivery delay and energy consumption.

Decentralized Coded Caching Scheme With Heterogeneous File Sizes

Coded caching is an effective technique to reduce peak network traffic load by exploiting coded multicasting opportunities among users with all possible requests. However, most existing works on coded caching design assume identical file sizes. In this paper, to fully exploit the coded multicasting opportunities in decentralized coded caching system with heterogenous file sizes, a file partitioning and grouping (FPG) scheme is proposed. In the FPG scheme, every file is divided into multiple subfiles, so that the subfiles of the same size from different files can be grouped into subfile groups. Based on the proposed FPG scheme, we formulate a placement optimized design problem to minimize the worst-case transmission load, in which we address how to allocate the limited cache sizes at users to different subfile groups. The optimal placement is found to be dependent on all file sizes and the cache size, as well as the number of subfile groups. The numerical results show that, the FPG scheme outperforms the existing decentralized coded caching schemes in the literature. Moreover, it can even approach the performance of the optimized centralized coded caching scheme when the number of files is smaller than the number of users.

Parloom: A New Low-Power Set-Associative Instruction Cache Architecture Utilizing Enhanced Counting Bloom Filter and Partial Tags

The cache system dissipates a significant amount of energy compared to the other memory components. This will be intensified if a cache is designed with a set-associative structure to improve the system performance because the parallel accesses to the entries of a set for tag comparisons lead to even more energy consumption. In this paper, a novel method is proposed as a combination of a counting Bloom filter and partial tags to mitigate the energy consumption of set-associative caches. This new hybrid method noticeably decreases the cache energy consumption especially in highly-associative instruction caches. In fact, it uses an enhanced counting Bloom filter to predict cache misses with a high accuracy as well as partial tags to decrease the overall cache size. This way, unnecessary tag comparisons can be prevented and therefore, the cache energy consumption is considerably reduced. Based on the simulation results, the proposed method provides the energy reduction from 22% to 31% for 4-way–32-way set-associative L1 caches bigger than 16[Formula: see text]kB running the MiBench programs. The improvements are attained with a negligible system performance degradation compared to the traditional cache system.

Deep Reinforcement Learning for Cooperative Content Caching in Vehicular Edge Computing and Networks

In this article, we propose a cooperative edge caching scheme, a new paradigm to jointly optimize the content placement and content delivery in the vehicular edge computing and networks, with the aid of the flexible trilateral cooperations among a macro-cell station, roadside units, and smart vehicles. We formulate the joint optimization problem as a double time-scale Markov decision process (DTS-MDP), based on the fact that the time-scale of content timeliness changes less frequently as compared to the vehicle mobility and network states during the content delivery process. At the beginning of the large time-scale, the content placement/updating decision can be obtained according to the content popularity, vehicle driving paths, and resource availability. On the small time-scale, the joint vehicle scheduling and bandwidth allocation scheme is designed to minimize the content access cost while satisfying the constraint on content delivery latency. To solve the long-term mixed integer linear programming (LT-MILP) problem, we propose a nature-inspired method based on the deep deterministic policy gradient (DDPG) framework to obtain a suboptimal solution with a low computation complexity. The simulation results demonstrate that the proposed cooperative caching system can reduce the system cost, as well as the content delivery latency, and improve content hit ratio, as compared to the noncooperative and random edge caching schemes.

CA-Based Detection of Coherence Exploiting Hardware Trojans

Nowadays, Hardware Trojan threats have become inevitable due to the growing complexities of Integrated Circuits (ICs) as well as the current trend of Intellectual Property (IP)-based hardware designs. An adversary can insert a Hardware Trojan during any of its life cycle phases — the design, fabrication or even at manufacturing phase. Once a Trojan is inserted into a system, it can cause an unwanted modification to system functionality which may degrade system performance or sometimes Trojans are implanted with the target to leak secret information. Once Trojans are implanted, they are hard to detect and impossible to remove from the system as they are already fabricated into the chip. In this paper, we propose three stealthy Trojan models which affect the coherence mechanism of Chip Multiprocessors’ (CMPs) cache system by arbitrarily modifying the cache block state which in turn may leave the cache line states as incoherent. We have evaluated the payload of such modeled Trojans and proposed a cellular automaton (CA)-based solution for detection of such Trojans.

A View Synthesis-Based 360° VR Caching System Over MEC-Enabled C-RAN

With the development of virtual reality (VR) technology, the future of VR systems is evolving from single-user wired connections to multi-user wireless connections. However, wireless online rendering and transmission incur extra processing and transmission latency, as well as higher bandwidth requirements. To meet the requirements of wireless VR applications and enhance the quality of the VR user experience, this paper designs a view synthesis-based 360° VR caching system over Cloud Radio Access Network (C-RAN), where both mobile edge computing (MEC) and hierarchical caching are supported. In the system, an MEC-Cache Server is deployed in the pooled Base band Units (BBU pool) and used for view synthesis and caching. In addition, the remote radio heads (RRHs) can also cache some video contents. If the requested content of a specific view is cached in the BBU pool or RRHs, or can be synthesized with the aid of the cached adjacent views, it is unnecessary to request the content from the remote VR video source server. Therefore, the transmission latency and backhaul traffic load for VR services can be decreased. We formulate a hierarchical collaborative caching problem aiming to minimize the transmission latency, which is proved NP-hard. To address the impractical expenses of the offline optimal method, an online MaxMinDistance caching algorithm with low complexity is proposed. Numerical simulation results demonstrate that the proposed caching strategy provides significantly improved cache hit rate, backhaul traffic load, transmission latency, and Quality of Experience (QoE) performances relative to conventional caching strategies.

Network Coding For Data Delivery in Caching at Edge: Concept, Model, and Algorithms

To deal with the explosive traffic demand, caching at the edge of mobile networks has been proposed to improve the quality of user experience. Presently, the edge caching systems are based on the direct and simple relay transmission, which is hard to achieve high system throughput and good content service. In such situations, network coding can effectively solve these problems. One problem arises in the application of network coding: how to form ‘X’ topology network. We formulate the ‘X’ topology formation problem as a function of system throughput and content service satisfaction degree. Then we propose a coalition game theory based algorithm to solve this NP-hard problem and prove the Nash stability of our coalition formation game. We evaluate our proposed algorithm through extensive simulations in different network environments. The experiment results demonstrate that our algorithm achieves nearly optimal performance and significantly outperforms the four widely-accepted baseline algorithms.

A Varying Processor Cache Sets Architecture

Any processor cache has three parameters capacity, line size and associativity. Usually all three are fixed at design time. Algorithms to have variable cache sets are proposed in literature. This paper proposes a method to have variable cache sets logically. The cache comes with fixed sets. The cache is visualized to have logically any number of sets greater than or equal to one. An algorithm for line placement/replacement is proposed in this paper for this model. The proposed model is simulated with SPEC2K benchmarks using Simplescalar Toolkit for two level inclusive set associative cache system. A power saving of 8.4% for L1 cache size 512x4, 17.58% for 1024x4 and 31.3% for 2048x4 is observed compared with traditional set associative cache of same size. A power saving of 7.53% compared with model proposed in literature for L1 size 512x4, 7.64% for 1024x4 and 7.645% for 2048x4 is observed. The L2 cache size is fixed at 2048x8. The average memory access time (AMAT) is found to degrade compared with conventional set associative cache by 19.63% for L1 size of 512x4, 24.68% for 1024x4 and 2048x4. (Abstract)

Modeling and Analysis of a Shared Edge Caching System for Connected Cars and Industrial IoT-Based Applications

The next revolution of industrial applications, known as smart industry or Industry 4.0, will rely on Internet of Things (IoT) to automate the monitoring, inspection, and control of industrial equipment and processes. In Industry 4.0, efficient content delivery is one of the fundamental challenges to be addressed. Nowadays, the promising solution for content delivery in smart industrial applications is the use of hierarchical caching systems at the network edge (5G small cells). This approach reduces the delay for content delivery and helps improve the performance of smart industrial applications. However, the caching management is a challenging and complex task, especially in those scenarios of shared storage resources on edge devices to support multiple concurrent applications (e.g., industrial, mobile users, and connected cars applications). In this article, we study the performance of a shared edge caching system for content delivery in smart industry and connected cars applications. To do so, we propose a mathematical framework to model the performance of a hierarchical shared edge caching system. The proposed mathematical framework considers the distinct content catalogs of the different applications (e.g., industrial and connected cars applications) and content request characteristics from industrial IoT devices and vehicles. Numerical results show that the performance of the shared edge caching system is sensitive to vehicular mobility (i.e., vehicular speed).