Significant Overhead Research Articles

OTOP: Optimized Transmission Power Controlled OBSS PD-Based Spatial Reuse for High Throughput in IEEE 802.11be WLANs

As wireless local area network (WLAN) devices become more prevalent in various environments, the IEEE 802.11ax standard (Wi-Fi 6), released in 2021, aims to improve the spectral efficiency. The overlapping basic service set (OBSS) packet detection (PD) based spatial reuse (SR) scheme is a main technology for this purpose which can increase channel access opportunities by adjusting parameters. However, the standard only provides boundary values for parameter adjustment and does not provide details on how to optimize the performance, which can lead to a performance degradation. The IEEE 802.11be standard (Wi-Fi 7), established to satisfy extremely high throughput (EHT) demands of next-generation applications, includes the OBSS PD based SR scheme and the coordinated spatial reuse (CSR) scheme to maximize area throughput in dense networks. The CSR scheme optimizes the transmission performance through coordination between access points (APs), which results in additional signaling overhead and protocol complexity. In this paper, an enhanced SR scheme named optimized transmission power based OBSS PD (OTOP) SR is proposed to achieve high throughput in dense networks. The OTOP scheme enhances the parameter control operation of OBSS PD based SR to take advantage of its distributed nature which keeps the signaling overhead and protocol complexity low. The OTOP scheme derives the optimal transmission power that can maximize the transmission success probability of WLAN devices based on stochastic geometry analysis. The simulation results show that the proposed OTOP scheme performs better than the existing SRGM and OBSS PD based SR schemes in terms of throughput, frame error rate, and fairness.

In-Network Computation for Large-Scale Federated Learning Over Wireless Edge Networks

Most conventional Federated Learning (FL) models are using a star network topology where all users aggregate their local models at a single server (e.g., a cloud server). That causes significant overhead in terms of both communications and computing at the server, delaying the training process, especially for large scale FL systems with straggling nodes. This paper proposes a novel edge network architecture that enables decentralizing the model aggregation process at the server, thereby significantly reducing the training delay for the whole FL network. Specifically, we design a highly-effective in-network computation framework (INC) consisting of a user scheduling mechanism, an in-network aggregation process (INA) which is designed for both primal- and primal-dual methods in distributed machine learning problems, and a network routing algorithm with theoretical performance bounds. The in-network aggregation process, which is implemented at edge nodes and cloud node, can adapt two typical methods to allow edge networks to effectively solve the distributed machine learning problems. Under the proposed INA, we then formulate a joint routing and resource optimization problem, aiming to minimize the aggregation latency. The problem turns out to be NP-hard, and thus we propose a polynomial time routing algorithm which can achieve near optimal performance with a theoretical bound. Simulation results showed that the proposed algorithm can achieve more than 99<inline-formula><tex-math notation="LaTeX">$\%$</tex-math></inline-formula> of the optimal solution and reduce the FL training latency, up to 5.6 times w.r.t other baselines. The proposed INC framework can not only help reduce the FL training latency but also significantly decrease cloud&#x2019;s traffic and computing overhead. By embedding the computing/aggregation tasks at the edge nodes and leveraging the multi-layer edge-network architecture, the INC framework can liberate FL from the star topology to enable large-scale FL.

Communication-Efficient Federated Learning on Non-IID Data Using Two-Step Knowledge Distillation

Federated learning (FL) has shown its great potential for achieving distributed intelligence in privacy-sensitive IoT. However, popular FL approaches such as FedAvg and its variants share model parameters among clients during the training process and thus cause significant communication overhead in IoT. Moreover, non-independent and identically distributed (non-IID) data across learning devices severely affect the convergence and speed of FL. To address these challenges, we propose a communication-efficient Federated Learning framework based on Two-step Knowledge Distillation, Fed2KD, which boosts the classification accuracy through privacy-preserving data generation while improving communication efficiency through a new knowledge distillation scheme empowered by an attention mechanism and metric learning. The generalization ability of Fed2KD is analyzed from the view of domain adaption. Extensive simulation experiments are conducted on Fashion-MNIST, CIFAR-10, and ImageNet datasets with various non-IID data distributions. The performance results show that Fed2KD can reduce the communication overhead and improve classification accuracy compared to FedAvg and its latest variants.

End-to-end Privacy Preserving Training and Inference for Air Pollution Forecasting with Data from Rival Fleets

Privacy-preserving machine learning (PPML) promises to train machine learning (ML) models by combining data spread across multiple data silos. Theoretically, secure multiparty computation (MPC) allows multiple data owners to train models on their joint data without revealing the data to each other. However, the prior implementations of this secure training using MPC have three limitations: they have only been evaluated on CNNs, and LSTMs have been ignored; fixed point approximations have affected training accuracies compared to training in floating point; and due to significant latency overheads of secure training via MPC, its relevance for practical tasks with streaming data remains unclear. The motivation of this work is to report our experience of addressing the practical problem of secure training and inference of models for urban sensing problems, e.g., traffic congestion estimation, or air pollution monitoring in large cities, where data can be contributed by rival fleet companies while balancing the privacy-accuracy trade-offs using MPC-based techniques.Our first contribution is to design a custom ML model for this task that can be efficiently trained with MPC within a desirable latency. In particular, we design a GCN-LSTM and securely train it on time-series sensor data for accurate forecasting, within 7 minutes per epoch. As our second contribution, we build an end-to-end system of private training and inference that provably matches the training accuracy of cleartext ML training. This work is the first to securely train a model with LSTM cells. Third, this trained model is kept secret-shared between the fleet companies and allows clients to make sensitive queries to this model while carefully handling potentially invalid queries. Our custom protocols allow clients to query predictions from privately trained models in milliseconds, all the while maintaining accuracy and cryptographic security.

Composite Preambles Based on Differential Phase Rotations for Grant-Free Random Access Systems

With the advantages of low signaling overhead and latency, grant-free random access (GFRA) becomes a promising technology for supporting massive machine-type communications (mMTC), but poses new challenges for active user detection (AUD) and channel estimation (CE), whose performance mainly depends on the preamble detection. In this paper, we design the composite preamble based on differential phase rotations by aggregating orthogonal Zadoff-Chu (ZC) sequences and multiple root ZC sequences with differential phase rotations to reduce the probability of preamble collisions, thereby improving the performance of AUD and CE. In particular, differential phase rotations extend the preamble set size so that users colliding in orthogonal sequences can be distinguished by phase rotations. In addition, it also reduces non-orthogonal interference and thus reduces CE errors. The preamble detection algorithm and CE scheme are proposed, along with the theoretical analysis of AUD and CE performance to verify the effectiveness of the designed preamble. In addition, the proposed preamble is extended to combine phase rotations with cyclic shifts to further enlarge the preamble set size with low non-orthogonality. Simulation results show that the proposed composite preamble outperforms existing preambles in terms of the probability of detection and CE accuracy.

Mobility Support for MIMO-NOMA User Clustering in Next-Generation Wireless Networks

Non-Orthogonal Multiple Access (NOMA) is a promising technology for future-generation wireless systems, with potential to contribute to the improvement of spectral efficiency. NOMA groups users into clusters, based on channel gain-difference. However, user mobility continuously changes the channel gain, which often requires re-clustering. In this paper, we study a set of re-clustering methods: arbitrary, one-by-one and Kuhn-Munkres assignment algorithm (KMAA), that expedite link re-establishment and keep the clusters interference-free, taking into account the mobility of users. The methods are applied to automatically dissociate identified users within clusters, when the gain-difference is lower than a given threshold, followed by re-association procedure, which integrates users into different clusters, maintaining an appropriate gain-difference. Experimental results show that the KMAA method improves efficiency and capacity through minimizing the number of re-clustering events, improving resource utilization, and lowering signaling overhead. Other sets of results highlight the throughput and outage probability gains of the KMAA method across a wide range of mobility scenarios. We also provide an analysis of the KMAA algorithm when applied to MIMO-NOMA, encompassing link resiliency and maintenance of average gain-difference, among users in clusters.

Learning agent-based security schema mitigating man-in-the-middle attacks in fog computing

&lt;div style="text-align: start;" align="center"&gt;&lt;span lang="EN-US"&gt;The fast emerging of internet of things (IoTs) has introduced fog computing as an intermediate layer between end-users and the cloud datacenters. Fog computing layer characterized by its closeness to end users for service provisioning than the cloud. However, security challenges are still a big concern in fog and cloud computing paradigms as well. In fog computing, one of the most destructive attacks is man-in-the-middle (MitM). Moreover, MitM attacks are hard to be detected since they performed passively on the network level. This paper proposes a MitM mitigation scheme in fog computing architecture. The proposal mapped the fog layer on software-defined network (SDN) architecture. The proposal integrated multi-path transmission control protocol (MPTCP), moving target defense (MTD) technique, and reinforcement learning agent (RL) in one framework that contributed significantly to improving the fog layer resources utilization and security. The proposed schema hardens the network reconnaissance and discovery, thus improved the network security against MitM attack. The evaluation framework was tested using a simulation environment on mininet, with the utilization of MPTCP kernel and Ryu SDN controller. The experimental results shows that the proposed schema maintained the network resiliency, improves resource utilization without adding significant overheads compared to the traditional transmission control protocol (TCP).&lt;/span&gt;&lt;/div&gt;

Dataset Condensation via Expert Subspace Projection.

The rapid growth in dataset sizes in modern deep learning has significantly increased data storage costs. Furthermore, the training and time costs for deep neural networks are generally proportional to the dataset size. Therefore, reducing the dataset size while maintaining model performance is an urgent research problem that needs to be addressed. Dataset condensation is a technique that aims to distill the original dataset into a much smaller synthetic dataset while maintaining downstream training performance on any agnostic neural network. Previous work has demonstrated that matching the training trajectory between the synthetic dataset and the original dataset is more effective than matching the instantaneous gradient, as it incorporates long-range information. Despite the effectiveness of trajectory matching, it suffers from complex gradient unrolling across iterations, which leads to significant memory and computation overhead. To address this issue, this paper proposes a novel approach called Expert Subspace Projection (ESP), which leverages long-range information while avoiding gradient unrolling. Instead of strictly enforcing the synthetic dataset's training trajectory to mimic that of the real dataset, ESP only constrains it to lie within the subspace spanned by the training trajectory of the real dataset. The memory-saving advantage offered by our method facilitates unbiased training on the complete set of synthetic images and seamless integration with other dataset condensation techniques. Through extensive experiments, we have demonstrated the effectiveness of our approach. Our method outperforms the trajectory matching method on CIFAR10 by 16.7% in the setting of 1 Image/Class, surpassing the previous state-of-the-art method by 3.2%.

An Innovative Architecture of DRAM PUF

Cryptographic solutions based on traditional authentication methods are susceptible to several attacks on secret keys. Dynamic random access memory (DRAM)-based physical unclonable functions (PUF) are described as promising security building blocks to enable cryptography and authentication services. PUFs frequently suffer reliability concerns since they are sensitive to both internal and exterior noises. The need for enhanced resistance and dependability results in significant additional overheads. In this work, we proposed a DRAM-PUF based on the inclusion of selective hardware features in the computation. The proposed solution offers a higher number of challenge-response pairs (CRP) without additional circuitry. An innovative structure of PUF is presented in the paper which offers a reliable challenge-response pair module for authentication and authorization based on the ubiquitous nature of memory without the need for an additional circuit. DRAM PUF utilized the random startup value initialized by a capacitor followed by a random number generator. Our proposed PUF shows a reliability of 99.46% with temperature variation, a reliability of 99.5% with supply voltage variation, a uniqueness of 49.46%, a bit aliasing of 46.875%, and a uniformity of 47.65%.

Tensor-Empowered LSTM for Communication-Efficient and Privacy-Enhanced Cognitive Federated Learning in Intelligent Transportation Systems

Multimedia cognitive computing as a revolutionary emerging concept of artificial intelligence emulating the reasoning process like human brains can facilitate the evolution of intelligent transportation systems (ITS) to be smarter, safer, and more efficient. Massive multimedia traffic big data is an important prerequisite for the success of cognitive computing in ITS. However, traditional data-centralized artificial intelligence approaches often face the problems of data islands and data famine due to concerns about data privacy and security. To this end, we propose the concept of cognitive federated learning leveraging federated learning as the learning paradigm for cognitive computing, which solves the preceding concerns by sharing updated models rather than raw data. Nevertheless, the exchange of numerous model parameters not only generates significant communication overhead but also suffers from the risk of privacy leakage due to inference attacks. This article aims to design a novel lightweight and privacy-enhanced cognitive federated learning architecture to facilitate the development of ITS. First, a privacy-enhanced model protection scheme with homomorphic encryption as the underlying technology is proposed to simultaneously defend against the inference attacks launched by external malicious attackers, honest-but-curious cognitive platforms, and internal participants. Furthermore, a novel tensor ring-block decomposition and its corresponding deep computation model converting the weight tensor into a set of matrices and third-order core tensors are proposed, which could reduce the communication overhead and storage requirements without compromising model performance. Experimental results on real-world datasets show that the proposed approach performs well.

DNNShifter: An efficient DNN pruning system for edge computing

Deep neural networks (DNNs) underpin many machine learning applications. Production quality DNN models achieve high inference accuracy by training millions of DNN parameters which has a significant resource footprint. This presents a challenge for resources operating at the extreme edge of the network, such as mobile and embedded devices that have limited computational and memory resources. To address this, models are pruned to create lightweight, more suitable variants for these devices. Existing pruning methods are unable to provide similar quality models compared to their unpruned counterparts without significant time costs and overheads or are limited to offline use cases. Our work rapidly derives suitable model variants while maintaining the accuracy of the original model. The model variants can be swapped quickly when system and network conditions change to match workload demand. This paper presents DNNShifter, an end-to-end DNN training, spatial pruning, and model switching system that addresses the challenges mentioned above. At the heart of DNNShifter is a novel methodology that prunes sparse models using structured pruning - combining the accuracy-preserving benefits of unstructured pruning with runtime performance improvements of structured pruning. The pruned model variants generated by DNNShifter are smaller in size and thus faster than dense and sparse model predecessors, making them suitable for inference at the edge while retaining near similar accuracy as of the original dense model. DNNShifter generates a portfolio of model variants that can be swiftly interchanged depending on operational conditions. DNNShifter produces pruned model variants up to 93x faster than conventional training methods. Compared to sparse models, the pruned model variants are up to 5.14x smaller and have a 1.67x inference latency speedup, with no compromise to sparse model accuracy. In addition, DNNShifter has up to 11.9x lower overhead for switching models and up to 3.8x lower memory utilisation than existing approaches. DNNShifter is available for public use from https://github.com/blessonvar/DNNShifter.

Fedcs: Efficient communication scheduling in decentralized federated learning

Decentralized federated learning is a training approach that prioritizes user data privacy protection, while also offering improved scalability and robustness. However, as the number of edge devices participating in training increases, a significant communication overhead arises among devices located in different geographical locations. Therefore, designing a well-thought-out gradient synchronization strategy is crucial for minimizing the overall communication overhead of training.To tackle this issue, this article introduces a 2D-Ring network structure based parameter synchronization strategy and a 2D-attention-based device placement algorithm, aiming to minimize communication overhead. The parameter synchronization strategy devises a two-layer circular communication architecture for the devices involved in training, thereby reducing the overall frequency of parameter synchronization in decentralized federated learning. By taking into account the total communication overhead and the device placement strategy, an optimization problem is formulated. Specifically, a 2D-attention neural network is constructed to optimize the device placement solution based on 2D-Ring network structure, leading to reduced communication overhead. Moreover, an evaluation model is designed to assess the communication overhead in a complex decentralized system during federated training. This enables precise determination of the total communication overhead throughout the training process, providing valuable insights for devising the device placement strategy. Extensive simulations confirm that the proposed approach achieves a substantial reductions of 55% and 64% in the total communication overhead for decentralized federated learning training with 50 and 100 devices, respectively.

Privacy and Data Balkanization: Circumventing the Barriers

The rapid growth in digital data forms the basis for a wide range of new services and research, e.g, large-scale medical studies. At the same time, increasingly restrictive privacy concerns and laws are leading to significant overhead in arranging for sharing or combining different data sets to obtain these benefits. For new applications, where the benefit of combined data is not yet clear, this overhead can inhibit organizations from even trying to determine whether they can mutually benefit from sharing their data. In this paper, we discuss techniques to overcome this difficulty by employing private information transfer to determine whether there is a benefit from sharing data, and whether there is room to negotiate acceptable prices. These techniques involve cryptographic protocols. While currently considered secure, these protocols are potentially vulnerable to the development of quantum technology, particularly for ensuring privacy over significant periods of time into the future. To mitigate this concern, we describe how developments in practical quantum technology can improve the security of these protocols.

Joint Resource Allocation and Drones Relay Selection for Large-Scale D2D Communication Underlaying Hybrid VLC/RF IoT Systems

Relay-aided Device-to-Device (D2D) communication combining visible light communication (VLC) with radio frequency (RF) is a promising paradigm in the internet of things (IoT). Static relay limits the flexibility and maintaining connectivity of relays in Hybrid VLC/RF IoT systems. By using a drone as a relay station, it is possible to avoid obstacles such as buildings and to communicate in a line-of-sight (LoS) environment, which naturally aligns with the requirement of VLC Systems. To further support the application of VLC in the IoT, subject to the challenges imposed by the constrained coverage, the lack of flexibility, poor reliability, and connectivity, drone relay-aided D2D communication appears on the horizon and can be cost-effectively deployed for the large-scale IoT. This paper proposes a joint resource allocation and drones relay selection scheme, aiming to maximize the D2D system sum rate while ensuring the quality of service (QoS) requirements for cellular users (CUs) and D2D users (DUs). First, we construct a two-phase coalitional game to tackle the resource allocation problem, which exploits the combination of VLC and RF, as well as incorporates a greedy strategy. After that, a distributed cooperative multi-agent reinforcement learning (MARL) algorithm, called WoLF policy hill-climbing (WoLF-PHC), is proposed to address the drones relay selection problem. Moreover, to further reduce the computational complexity, we propose a lightweight neighbor–agent-based WoLF-PHC algorithm, which only utilizes historical information of neighboring DUs. Finally, we provide an in-depth theoretical analysis of the proposed schemes in terms of complexity and signaling overhead. Simulation results illustrate that the proposed schemes can effectively improve the system performance in terms of the sum rate and outage probability with respect to other outstanding algorithms.

Exploring temporal community evolution: algorithmic approaches and parallel optimization for dynamic community detection

Dynamic (temporal) graphs are a convenient mathematical abstraction for many practical complex systems including social contacts, business transactions, and computer communications. Community discovery is an extensively used graph analysis kernel with rich literature for static graphs. However, community discovery in a dynamic setting is challenging for two specific reasons. Firstly, the notion of temporal community lacks a widely accepted formalization, and only limited work exists on understanding how communities emerge over time. Secondly, the added temporal dimension along with the sheer size of modern graph data necessitates new scalable algorithms. In this paper, we investigate how communities evolve over time based on several graph metrics under a temporal formalization. We compare six different algorithmic approaches for dynamic community detection for their quality and runtime. We identify that a vertex-centric (local) optimization method works as efficiently as the classical modularity-based methods. To its advantage, such local computation allows for the efficient design of parallel algorithms without incurring a significant parallel overhead. Based on this insight, we design a shared-memory parallel algorithm DyComPar, which demonstrates between 4 and 18 fold speed-up on a multi-core machine with 20 threads, for several real-world and synthetic graphs from different domains.

Improving Worst-case TSN Communication Times of Large Sensor Data Samples by Exploiting Synchronization

Higher levels of automated driving also require a more sophisticated environmental perception. Therefore, an increasing number of sensors transmit their data samples as frame bursts to other applications for further processing. As a vehicle has to react to its environment in time, such data is subject to safety-critical latency constraints. To keep up with the resulting data rates, there is an ongoing transition to a Time-Sensitive Networking (TSN)-based communication backbone. However, the use of TSN-related industry standards does not match the automotive requirements of large timely sensor data transmission, nor it offers benefits on time-critical transmissions of single control data packets. By using the full data rate of prioritized IEEE 802.1Q Ethernet, giving time guarantees on large data samples is possible, but with strongly degraded results due to data collision. Resolving such collisions with time-aware shaping comes with significant overhead. Hence, rather than optimizing the parameters of the existing protocol, we propose a system design that synchronizes the transmission times of sensor data samples. This limits network protocol complexity and hardware requirements by avoiding tight time synchronization and time-aware shaping. We demonstrate that individual sensor data samples are transmitted without significant interference, exclusively at full Ethernet data rate. We provide a synchronous event model together with a straightforward response time analysis for synchronous multi-frame sample transmissions. The results show that worst-case latencies of such sample communication, in contrast to non-synchronized approaches, are close to their theoretical minimum as well as to simulative results while keeping the overall network utilization high.

IOSR: Improving I/O Efficiency for Memory Swapping on Mobile Devices Via Scheduling and Reshaping

Mobile systems and applications are becoming increasingly feature-rich and powerful, which constantly suffer from memory pressure, especially for devices equipped with limited DRAM. Swapping inactive DRAM pages to the storage device is a promising solution to extend the physical memory. However, existing mobile devices usually adopt flash memory as the storage device, where swapping DRAM pages to flash memory may introduce significant performance overhead. In this paper, we first conduct an in-depth analysis of the I/O characteristics of the flash-based memory swapping, including the I/O interference and swap I/O randomness in swap subsystem. Then an I/O efficiency optimization framework for memory swapping (IOSR) is proposed to enhance the performance of flash-based memory swapping for mobile devices. IOSR consists of two methods: swap I/O scheduling (SIOS) and swap I/O pattern reshaping (SIOR). SIOS is designed to schedule the swap I/O to reduce interference with other processes I/Os. SIOR is designed to reshape the swap I/O pattern with process-oriented swap slot allocation and adaptive granularity swap read-ahead. IOSR is implemented on Google Pixel 4. Experimental results show that IOSR reduces the application switching time by 31.7% and improves the swap-in bandwidth by 35.5% on average compared to the state-of-the-art.

CIM: A Novel Clustering-based Energy-Efficient Data Imputation Method for Human Activity Recognition

Human activity recognition (HAR) is an important component in a number of health applications, including rehabilitation, Parkinson’s disease, daily activity monitoring, and fitness monitoring. State-of-the-art HAR approaches use multiple sensors on the body to accurately identify activities at runtime. These approaches typically assume that data from all sensors are available for runtime activity recognition. However, data from one or more sensors may be unavailable due to malfunction, energy constraints, or communication challenges between the sensors. Missing data can lead to significant degradation in the accuracy, thus affecting quality of service to users. A common approach for handling missing data is to train classifiers or sensor data recovery algorithms for each combination of missing sensors. However, this results in significant memory and energy overhead on resource-constrained wearable devices. In strong contrast to prior approaches, this paper presents a clustering-based approach (CIM) to impute missing data at runtime. We first define a set of possible clusters and representative data patterns for each sensor in HAR. Then, we create and store a mapping between clusters across sensors. At runtime, when data from a sensor are missing, we utilize the stored mapping table to obtain most likely cluster for the missing sensor. The representative window for the identified cluster is then used as imputation to perform activity classification. We also provide a method to obtain imputation-aware activity prediction sets to handle uncertainty in data when using imputation. Experiments on three HAR datasets show that CIM achieves accuracy within 10% of a baseline without missing data for one missing sensor when providing single activity labels. The accuracy gap drops to less than 1% with imputation-aware classification. Measurements on a low-power processor show that CIM achieves close to 100% energy savings compared to state-of-the-art generative approaches.

DTRL: Decision Tree-based Multi-Objective Reinforcement Learning for Runtime Task Scheduling in Domain-Specific System-on-Chips

Domain-specific systems-on-chip (DSSoCs) combine general-purpose processors and specialized hardware accelerators to improve performance and energy efficiency for a specific domain. The optimal allocation of tasks to processing elements (PEs) with minimal runtime overheads is crucial to achieving this potential. However, this problem remains challenging as prior approaches suffer from non-optimal scheduling decisions or significant runtime overheads. Moreover, existing techniques focus on a single optimization objective, such as maximizing performance. This work proposes DTRL, a decision-tree-based multi-objective reinforcement learning technique for runtime task scheduling in DSSoCs. DTRL trains a single global differentiable decision tree (DDT) policy that covers the entire objective space quantified by a preference vector. Our extensive experimental evaluations using our novel reinforcement learning environment demonstrate that DTRL captures the trade-off between execution time and power consumption, thereby generating a Pareto set of solutions using a single policy. Furthermore, comparison with state-of-the-art heuristic–, optimization–, and machine learning-based schedulers shows that DTRL achieves up to 9× higher performance and up to 3.08× reduction in energy consumption. The trained DDT policy achieves 120 ns inference latency on Xilinx Zynq ZCU102 FPGA at 1.2 GHz, resulting in negligible runtime overheads. Evaluation on the same hardware shows that DTRL achieves up to 16% higher performance than a state-of-the-art heuristic scheduler.

Object Fingerprint Cache for Heterogeneous Memory System

Heterogeneous memory systems promise to provide both large storage capacity and high performance. Prior DRAM cache systems have metadata scalability issue and suffer from low cache hit rate, low DRAM space utilization, and significant data migration overhead. We observe that instances of an object type exhibit stable and predictable memory access patterns in its constituent cachelines and these patterns are referred to as object fingerprints. We propose the hardware-assisted cache that manages DRAM at the object type level and fetches data block at the granularity of a cacheline, by exploiting object fingerprint. To address its design challenges, we first present a software-hardware co-design to convey the software information to hardware. Second, we design multiple granularity sector caches that can be dynamically adjusted to adapt to changing behaviors and improve DRAM cache utilization. To address the challenges of large metadata storage overhead, we propose to bound possible sizes for each sector cache. Experimental results show our designs improve DRAM cache hit rate by 21.6%, boost IPC by 19.8%, and reduce data migration traffic by 51.6% on average, compared with state-of-art DRAM caches. More importantly, our online object fingerprint learning method is 2.3% inferior to the offline one in terms of IPC.