Local Cache Research Articles

Reconciling Earlier Snapshot Time with Local Cache for Optimal Performance under Transactional Causal Consistency

Currently, Transactional Causal Consistency (TCC) has succeeded in removing some anomalies discovered in eventual consistency, diminishing the synchronization overhead consumed by strong consistency, and advocating interactive read-write transactions. However, the popular design in maximizing snapshot time for data freshness weighs against read-write interaction especially under read heavy overloads. Moreover, lower end-user latency and higher throughput in reading becomes paramount desiderata in deploying web services such as social networks among many data-centers. To this purpose, this article proposes earlier snapshot time and client-side local data-center cache for read-heavy transaction to boost lower response time and higher throughput. By the former, our new causal dependencies tracking mechanism can precisely initialize transactions to get the version values. For the latter, a large number of remote collections are eliminated since non-replica data is cached in the local data center. On the other hand, our system benefits from partial replication and one dependency-tracked timestamp to boost resource efficiency and scalability. Practically, our performance evaluation reveals that our design garners 6.73x lower response time and 1.76x higher throughput than the conventional maximized coordinating snapshot.

Optimizing Coherence Traffic in Manycore Processors Using Closed-Form Caching/Home Agent Mappings

Manycore processors feature a high number of general-purpose cores designed to work in a multithreaded fashion. Recent manycore processors are kept coherent using scalable distributed directories. A paramount example is the Intel Mesh interconnect, which consists of a network-on-chip interconnecting “tiles”, each of which contains computation cores, local caches, and coherence masters. The distributed coherence subsystem must be queried for every out-of-tile access, imposing an overhead on memory latency. This paper studies the physical layout of an Intel Knights Landing processor, with a particular focus on the coherence subsystem, and uncovers the pseudo-random mapping function of physical memory blocks across the pieces of the distributed directory. Leveraging this knowledge, candidate optimizations to improve memory latency through the minimization of coherence traffic are studied. Although these optimizations do improve memory throughput, ultimately this does not translate into performance gains due to inherent overheads stemming from the computational complexity of the mapping functions.

Integrated VLC/RF Wireless Technologies for Reliable Content Caching System in Vehicular Networks

In a vehicular communications environment, the need for information sharing, entertainment, and multimedia will increase, leading to congestion of backhaul networks. The major challenge of this network is latency and resource limitations. Proactive caching can be obtained from local caches rather than from remote servers, which can avoid long delays resulting from limited backhaul capacity and resources. Therefore, proactive caching reduces latency and improves the quality of services. Determining which files should be cached in memory is a critical issue. The paper proposes various placement schemes for caching at the vehicle and mobile base station (MBS) layers. The optimal caching place represents better user demand and latency based on the demand and mobility model. Non-cooperative and cooperative vehicles with vehicle-to-vehicle communication (V2V) are included. In both models, the high data rate of visible light communication (VLC) in the vehicular networks tends to enhance the caching capability. We also illustrate a sub-optimal algorithm to solve optimization problems compared to the optimal brute force solver and the sub-optimal genetic algorithm. Numerical results show that proactive caching schemes have a significant gain in system performance. The analysis shows that VLC improves proactive caching performance when there are less than 25 vehicles. The joint proactive caching scheme works better as the number of vehicles increases. The V2V proactive caching is useful when vehicles per platoon are ≥ 10

DCoL: Distributed Collaborative Learning for Proactive Content Caching at Edge Networks

Caching popular content at the network edge, such as roadside units (RSUs), is a promising solution that enhances the user’s quality-of-experience (QoE) and reduces network traffic. In this regard, the most challenging issue is to correctly predict the future popularity of contents and effectively store them in the cache of edge nodes. Thus, in this paper, we propose a distributed proactive caching scheme at the edge to optimize the content retrieval cost and improve the QoE of the mobile users. This proactive content caching scheme, namely Distributed Collaborative Learning (DCoL), is a non-parametric content popularity prediction mechanism in a distributed setting. Next, we show the advantage of DCoL as two folds: (i) it leverages distributed content popularity information to develop local content caching strategy, and (ii) it exploits the regional database using the long short-term memory (LSTM)-based prediction model to capture the dependency between requested contents. Simulation results using real datasets demonstrate that our scheme yields 8.9% and 18% gains, respectively, in terms of the cache hit efficiency and content retrieval cost, compared with a competitive centralized baseline, and outperforms other traditional caching strategies.

OmpSs@FPGA framework for high performance FPGA computing

This article presents the new features of the OmpSs@FPGA framework. OmpSs is a data-flow programming model that supports task nesting and dependencies to target asynchronous parallelism and heterogeneity. OmpSs@FPGA is the extension of the programming model addressed specifically to FPGAs. OmpSs environment is built on top of Mercurium source to source compiler and Nanos++ runtime system. To address FPGA specifics Mercurium compiler implements several FPGA related features as local variable caching, wide memory accesses or accelerator replication. In addition, part of the Nanos++ runtime has been ported to hardware. Driven by the compiler this new hardware runtime adds new features to FPGA codes, such as task creation and dependence management, providing both performance increases and ease of programming. To demonstrate these new capabilities, different high performance benchmarks have been evaluated over different FPGA platforms using the OmpSs programming model. The results demonstrate that programs that use the OmpSs programming model achieve very competitive performance with low to moderate porting effort compared to other FPGA implementations.

Runtime Randomized Relocation of Crypto Libraries for Mitigating Cache Attacks

Crypto libraries such as OpenSSL and Libgcrypt are essential building blocks for implementing secure cloud services. Unfortunately, these libraries are subject to cache side-channel attacks, which are more devastating in cloud environments where inevitable cache contention among different tenants occurs. Previous approaches for mitigating cache side-channel attacks have limitations in terms of the deployability and security; these hinder utilization in cloud services. In this paper, we propose an R2-relocator, a novel library protection technique based on moving target defence. When injected into a running process, the R2-relocator performs randomized relocation of the library during runtime. By doing this, it transforms a vulnerable crypto library into one that randomly changes its memory (cache) location, thereby preventing the delivery of cache side-channel attacks against the library. The proposed technique achieves robust protection against cache side-channel attacks for all crypto libraries, even those containing unpatched critical vulnerabilities, without the need for reconfiguration of the library. Extensive evaluations of security, performance, and deployability of the R2-relocator demonstrate its effectiveness for secure cloud services.

An FPGA‐Based Convolutional Neural Network Coprocessor

In this paper, an FPGA‐based convolutional neural network coprocessor is proposed. The coprocessor has a 1D convolutional computation unit PE in row stationary (RS) streaming mode and a 3D convolutional computation unit PE chain in pulsating array structure. The coprocessor can flexibly control the number of PE array openings according to the number of output channels of the convolutional layer. In this paper, we design a storage system with multilevel cache, and the global cache uses multiple broadcasts to distribute data to local caches and propose an image segmentation method that is compatible with the hardware architecture. The proposed coprocessor implements the convolutional and pooling layers of the VGG16 neural network model, in which the activation value, weight value, and bias value are quantized using 16‐bit fixed‐point quantization, with a peak computational performance of 316.0 GOP/s and an average computational performance of 62.54 GOP/s at a clock frequency of 200 MHz and a power consumption of about 9.25 W.

Scaling Performance Analysis and Optimization Based on the Node Spatial Distribution in Mobile Content‐Centric Networks

Content‐centric networks (CCNs) have become a promising technology for relieving the increasing wireless traffic demands. In this paper, we explore the scaling performance of mobile content‐centric networks based on the nonuniform spatial distribution of nodes, where each node moves around its own home point and requests the desired content according to a Zipf distribution. We assume each mobile node is equipped with a finite local cache, which is applied to cache contents following a static cache allocation scheme. According to the nonuniform spatial distribution of cache‐enabled nodes, we introduce two kinds of clustered models, i.e., the clustered grid model and the clustered random model. In each clustered model, we analyze throughput and delay performance when the number of nodes goes infinity by means of the proposed cell‐partition scheduling scheme and the distributed multihop routing scheme. We show that the node mobility degree and the clustering behavior play the fundamental roles in the aforementioned asymptotic performance. Finally, we study the optimal cache allocation problem in the two kinds of clustered models. Our findings provide a guidance for developing the optimal caching scheme. We further perform the numerical simulations to validate the theoretical scaling laws.

Node Value and Content Popularity‐Based Caching Strategy for Massive VANETs

The high‐speed dynamic environment and massive information transmitted via wireless communications in the vehicular ad hoc networks (VANETs) pose a great challenge to privacy and security. To overcome this issue, use of the content‐centric networking (CCN) provides a potential and practical solution. In‐network caching is a main feature for future smart cities, in which the content is mainly placed in network nodes. Therefore, how to effectively select the cache locality and cache content is essential to improve the overall network performance, which is an inevitable trend. With these observations, this article proposes a caching strategy based on the node value and content popularity (NVCP) for the massive VANET scenario. In the proposed NVCP scheme, different from the traditional caching strategies, we evaluate the node value from three aspects: the connectivity, intermediary, and eigenvector centralities, synthetically, since the content with different types of popularity is placed in nodes with different values, resulting in the redundancy deterioration and diversity improvement for the content. The proposed caching strategy is evaluated by the stochastic network topology with multifactors, which provides different impacts on the system performance. Simulation results show that the NVCP outperforms the traditional cache strategies for 6G‐CCN in terms of the cache hit ratio, average hop count, and transmission latency. Moreover, placing the content in the neighbor nodes is also introduced to further improve the utilization of the cache space and achieve better cache performance.

Communication Optimization Schemes for Accelerating Distributed Deep Learning Systems

In a distributed deep learning system, a parameter server and workers must communicate to exchange gradients and parameters, and the communication cost increases as the number of workers increases. This paper presents a communication data optimization scheme to mitigate the decrease in throughput due to communication performance bottlenecks in distributed deep learning. To optimize communication, we propose two methods. The first is a layer dropping scheme to reduce communication data. The layer dropping scheme we propose compares the representative values of each hidden layer with a threshold value. Furthermore, to guarantee the training accuracy, we store the gradients that are not transmitted to the parameter server in the worker’s local cache. When the value of gradients stored in the worker’s local cache is greater than the threshold, the gradients stored in the worker’s local cache are transmitted to the parameter server. The second is an efficient threshold selection method. Our threshold selection method computes the threshold by replacing the gradients with the L1 norm of each hidden layer. Our data optimization scheme reduces the communication time by about 81% and the total training time by about 70% in a 56 Gbit network environment.

Communications-Caching-Computing Resource Allocation for Bidirectional Data Computation in Mobile Edge Networks

A novel bidirectional computation task model has emerged as an important use case of 5G. For example, interactive AR/VR gaming service needs to render the live scene by jointly computing user features such as 3D positions and video data generated from the Internet. In this article, we consider the bidirectional computation task model, where each task is served via three mechanisms, i.e., local computing with local caching, local computing without local caching, and computing at the mobile edge computing server. To minimize the average bandwidth, we formulate the joint caching and computing optimization problem under the latency, cache size and average power constraints. In the homogeneous scenario, we derive the optimal policy and analytical expression for the minimum bandwidth. In the heterogeneous scenario, to reduce the computation complexity of the NP-hard problem, we relax some constraints of the original problem and propose a Lagrangian relaxation (LR) suboptimal solution, which may be infeasible. We then reformulate the original problem as an auxiliary problem based on the LR solution and solve this via Concave-Convex Procedure (CCCP), which outputs feasible local optimal solution. Simulation has shown that LR-based algorithms outperform the baselines including greedy and CCCP algorithms in the bandwidth performance and time efficiency.

On Coded Caching With Private Demands

Caching is an efficient way to reduce network traffic congestion during peak hours by storing some content at the user’s local cache memory without knowledge of later demands. For the shared-link caching model, Maddah-Ali and Niesen (MAN) proposed a two-phase ( placement and delivery ) coded caching strategy, which is order optimal within a constant factor. However, in the MAN coded caching scheme, each user can obtain the information about the demands of other users, i.e., the MAN coded caching scheme is inherently prone to tampering and spying the activity/demands of other users. In this paper, we formulate an information-theoretic shared-link caching model with private demands, where there are ${\mathsf K}$ cache-aided users (which can cache up to ${\mathsf M}$ files) connected to a central server with access to ${\mathsf N}$ files. Each user requests ${\mathsf L}$ files. Our objective is to design a two-phase private caching scheme with minimum load while preserving the information-theoretic privacy of the demands of each user with respect to other users. A trivial solution is the uncoded caching scheme which lets each user recover all the ${\mathsf N}$ files, referred to as baseline scheme . For this problem we propose two novel schemes which achieve the information-theoretic privacy of the users’ demands while also achieving a non-trivial caching gain over the baseline scheme. The general underlying idea is to satisfy the users’ requests by generating a set of coded multicast messages that is symmetric with respect to the library files, such that for each user k, the mutual information between these messages and the demands of all other users given the cache content and the demand of user k is zero. In the first scheme, referred to as virtual-user scheme, we introduce a number of virtual users such that each ${\mathsf L}$ -subset of files is demanded by ${\mathsf K}$ real or virtual (effective) users and use the MAN delivery to generate multicast messages. From the viewpoint of each user, the set of multicast messages is symmetric over all files even if each single multicast message is not. This scheme incurs in an extremely large sub-packetization. Then, we propose a second scheme, referred to as MDS-based scheme, based on a novel MDS-coded cache placement. In this case, we generate multicast messages where each multicast message contains one MDS-coded symbol from each file in the library and thus is again symmetric over all the files from the viewpoint of each user. The sub-packetization level of the MDS-based scheme is exponentially smaller than that needed by the virtual-user scheme. Compared with the existing shared-link coded caching converse bounds without privacy, the virtual-user scheme is proved to be order optimal with a constant factor when ${\mathsf N}\leq {\mathsf L} {\mathsf K} $ , or when ${\mathsf N}\geq {\mathsf L} {\mathsf K} $ and ${\mathsf M}\geq {\mathsf N}/ {\mathsf K}$ . In addition, when ${\mathsf M}\geq {\mathsf N}/2$ , both of the virtual-user scheme and the MDS-based scheme are order optimal within a factor of 2.

Optimizing performance of GATK workflows using Apache Arrow In-Memory data framework

BackgroundImmense improvements in sequencing technologies enable producing large amounts of high throughput and cost effective next-generation sequencing (NGS) data. This data needs to be processed efficiently for further downstream analyses. Computing systems need this large amounts of data closer to the processor (with low latency) for fast and efficient processing. However, existing workflows depend heavily on disk storage and access, to process this data incurs huge disk I/O overheads. Previously, due to the cost, volatility and other physical constraints of DRAM memory, it was not feasible to place large amounts of working data sets in memory. However, recent developments in storage-class memory and non-volatile memory technologies have enabled computing systems to place huge data in memory to process it directly from memory to avoid disk I/O bottlenecks. To exploit the benefits of such memory systems efficiently, proper formatted data placement in memory and its high throughput access is necessary by avoiding (de)-serialization and copy overheads in between processes. For this purpose, we use the newly developed Apache Arrow, a cross-language development framework that provides language-independent columnar in-memory data format for efficient in-memory big data analytics. This allows genomics applications developed in different programming languages to communicate in-memory without having to access disk storage and avoiding (de)-serialization and copy overheads.ImplementationWe integrate Apache Arrow in-memory based Sequence Alignment/Map (SAM) format and its shared memory objects store library in widely used genomics high throughput data processing applications like BWA-MEM, Picard and GATK to allow in-memory communication between these applications. In addition, this also allows us to exploit the cache locality of tabular data and parallel processing capabilities through shared memory objects.ResultsOur implementation shows that adopting in-memory SAM representation in genomics high throughput data processing applications results in better system resource utilization, low number of memory accesses due to high cache locality exploitation and parallel scalability due to shared memory objects. Our implementation focuses on the GATK best practices recommended workflows for germline analysis on whole genome sequencing (WGS) and whole exome sequencing (WES) data sets. We compare a number of existing in-memory data placing and sharing techniques like ramDisk and Unix pipes to show how columnar in-memory data representation outperforms both. We achieve a speedup of 4.85x and 4.76x for WGS and WES data, respectively, in overall execution time of variant calling workflows. Similarly, a speedup of 1.45x and 1.27x for these data sets, respectively, is achieved, as compared to the second fastest workflow. In some individual tools, particularly in sorting, duplicates removal and base quality score recalibration the speedup is even more promising.AvailabilityThe code and scripts used in our experiments are available in both container and repository form at: https://github.com/abs-tudelft/ArrowSAM.

An Energy-Efficient Cache Localization Technique for D2D Communication in IoT Environment

In the last few years, we have witnessed the cache localization as one of the most challenging problems for device-to-device (D2D) communication in the IoT environment. It has been a major performance bottleneck due to cache localization problem in D2D communication as there are advancements in cellular technology, especially in 5G base stations (BSs) deployment around the globe. It is due to the fact that with an increase in the enormous amount of the number of users and devices, there has been an increase in the demands of service availability within a fraction of seconds by the end users. It results in an increase in burden on the existing network infrastructure with respect to Quality of Service (QoS) and Quality of Experience (QoE) provisions to the end users and service providers. However, caching the most popular content on the user equipments (UE's) can resolve the aforementioned problems. Motivated from these facts, in this article, we propose a model to address the problem of the cache localization decision making. In the proposed scheme, first, we collected the data set traces to predict the cache locations. Then, we predicted the locations where the user can cache the most accessed content using machine learning classification models. The classification models used in the proposed solution are decision tree and random forest. The metrics used for evaluation of the results obtained are access delay and energy consumption of the UEs. On comparing the proposal with the other existing state-of-the-art models, we observed that the random forest model yields higher accuracy as compared to other existing models. Also, we have observed that the access delay is maximum at the user's end when contents are shared with the gateway.

An efficient content placement scheme based on normalized node degree in content centric networking

Content-centric networking (CCN) becomes an imminent Internet architecture that emphasizes the data-centric approach for information retrieval instead of searching for the hosts in the network. CCN offers data (content) caching and its distribution capabilities to reduce average latency, the load of data servers, and improve network bandwidth utilization during content delivery. In this direction, the performance of node degree centrality and distance-based autonomous caching strategies have been investigated. Then, a novel content caching strategy called content placement based on normalized node degree and distance (CPNDD) is proposed. The CPNDD scheme considers content provider distance and node degree centrality parameters together to select the optimal cache locations for comprehensive utilization of the available caching capacities and a further reduction in average latency during content retrieval. The weights for these caching parameters have been determined via widespread simulations on the Abilene networks. The proposed caching scheme is implemented and tested in the simulated environment of ndnSIM and compared with peer competing schemes in CCN. The execution outcomes are examined for different caching capacities (50 and 100), Zipf popularity skewness factors (0.7 and 1.0) and request rates (50/s and 100/s). The simulation executions illustrate that CPNDD caching strategy escalates the cache hit probability and hop-reduction ratio up-to $$8\%$$ and $$9\%$$ as compared to existing schemes. Hence, the proposed content placement scheme improves the performance of CCN networks considerably and makes it suitable for the applications of Industry 4.0.

Cooperation Based Proactive Caching in Multi-Tier Cellular Networks

The limited caching capacity of the local cache enabled Base station (BS) decreases the cache hit ratio (CHR) and user satisfaction ratio (USR). However, Cache enabled multi-tier cellular networks have been presented as a promising candidate for fifth generation networks to achieve higher CHR and USR through densification of networks. In addition to this, the cooperation among the BSs of various tiers for cached data transfer, intensify its significance many folds. Therefore, in this paper, we consider maximization of CHR and USR in a multi-tier cellular network. We formulate a CHR and USR problem for multi-tier cellular networks while putting major constraints on caching space of BSs of each tier. The unsupervised learning algorithms such as K-mean clustering and collaborative filtering have been used for clustering the similar BSs in each tier and estimating the content popularity respectively. A novel scheme such as cluster average popularity based collaborative filtering (CAP-CF) algorithm is employed to cache popular data and hence maximizing the CHR in each tier. Similarly, two novel methods such as intra-tier and cross-tier cooperation (ITCTC) and modified ITCTC algorithms have been employed in order to optimize the USR. Simulations results witness, that the proposed schemes yield significant performance in terms of average cache hit ratio and user satisfaction ratio compared to other conventional approaches.

Geocaching in education – a review of international experiences Part 3.Organisation of classes

AbstractThis article discusses the organization and conduct of educational activities using geocaching. In the scientific literature, the organization of geocaching classes is divided into three main stages: preparation, field work and evaluation. The preparation stage includes issues such as the role of the educator, the scenario itself, the duration of the classes, the length and course of the route, the number and location of caches, the coordinates and type of caches, potential descriptions of caches or the preparation of hints as well as the content of caches, necessary aids and preparing the students. In terms of the implementation of the fieldwork, the following issues need to be addressed: the role of the educator, introduction to classes, division of participants into teams, locating and working with caches as well as a summary of the activities. The evaluation stage should include an evaluation of the participants’ performance, the activities/thematic trails by the participants and the educational effect by the organizers. Additionally, other educational possibilities for using geocaching, such as tasks based on internet geocaching services, the use of existing caches and the establishment of caches by students are also briefly discussed.Due to the universality of the recommendations presented in the literature, educational geocaching can be implemented in every country, including Poland, both in formal and informal education, e.g. in the State Forests, national parks and landscape parks. Although this innovative teaching method involves a lot of initial work, its positive educational and social effects more than compensate for the invested time.

Social-Aware Proactive Content Caching and Sharing in Multi-Access Edge Networks

The ever-increasing data explosion, particular mobile video traffic, increases the backhaul load and makes it difficult for the centralized cloud to meet the requirements of various services. Accordingly, proactive caching at mobile devices gains more attentions, which is envisioned as a promising technology to relieve the backhaul traffic and cater to diverse quality of service in multi-access edge networks. In this article, we study the problem of proactive caching and content sharing for mobile devices in multi-access edge networks. Since users’ mobility behaviors and social impacts on content preference distort the performance of content requests, a long-term strategy of proactive caching is proposed to minimize the sum of communication costs to get the requested contents. By taking the mutual interaction of devices into account, we design a local cooperative utility function that is used to derive a distributed proactive content caching algorithm using game theory. Theoretical analysis is provided to confirm the stability and optimality of the proposed algorithm. Furthermore, simulation results show that the proposed algorithm converges efficiently and, most importantly, it outperforms compared to no local cooperation caching scheme, non-social aware caching scheme and randomly caching scheme for various kinds of user behaviors and device caching capacity.

MIPLIBing: Seamless Benchmarking of Mathematical Optimization Problems and Metadata Extensions

Public libraries of problems such as Mixed Integer Programming Library (MIPLIB) are fundamental to creating a common benchmark for measuring algorithmic advances across mathematical optimization solvers. They also often provide metadata on problem structure, hardness with respect to state-of-the-art solvers, and solutions with the best objective function value on record. In this short paper, we discuss some ways in which such metadata can be leveraged to create a seamless testing experience. In particular, we present MIPLIBing: a Python library that automatically downloads queried subsets from the current versions of MIPLIB, MINLPLib, and QPLIB, provides a centralized local cache across projects, and tracks the best solution values and bounds on record for each problem. While inspired by similar use cases from other areas, we reflect on the specific needs of mathematical optimization and discuss opportunities to extend benchmark sets to facilitate experimentation with different model structures.

Gini impurity based NDN cache pollution attack defence mechanism

Name Data Network (NDN) is a futuristic Internet architecture that delivers content based on strategically chosen names and caches content on intermediate routers in the content-delivery path, which is used to fulfill future requests for the same data. In-network caching significantly improves network throughput but also makes NDN vulnerable to revamp security attacks such as cache poisoning, cache pollution, and cache snooping. In this paper, we focused on content pollution attack, in which the attacker disrupts the cache locality by intelligently crafted fake requests. This paper introduce monitor metrics for quantitative anomaly detection of requests and proposed a scheme based on Gini Impurity to detect network under cache pollution attack. The paper also proposed an Interest throttling mechanism based on trust to reinforce network under cache pollution attacks. The proposed scheme is evaluated through various experiments on NdnSIM simulator.