Heterogeneous Workloads Research Articles

Adaptive and Scalable Database Management with Machine Learning Integration: A PostgreSQL Case Study

The increasing complexity of managing modern database systems, particularly in terms of optimizing query performance for large datasets, presents significant challenges that traditional methods often fail to address. This paper proposes a comprehensive framework for integrating advanced machine learning (ML) models within the architecture of a database management system (DBMS), with a specific focus on PostgreSQL. Our approach leverages a combination of supervised and unsupervised learning techniques to predict query execution times, optimize performance, and dynamically manage workloads. Unlike existing solutions that address specific optimization tasks in isolation, our framework provides a unified platform that supports real-time model inference and automatic database configuration adjustments based on workload patterns. A key contribution of our work is the integration of ML capabilities directly into the DBMS engine, enabling seamless interaction between the ML models and the query optimization process. This integration allows for the automatic retraining of models and dynamic workload management, resulting in substantial improvements in both query response times and overall system throughput. Our evaluations using the Transaction Processing Performance Council Decision Support (TPC-DS) benchmark dataset at scale factors of 100 GB, 1 TB, and 10 TB demonstrate a reduction of up to 42% in query execution times and a 74% improvement in throughput compared with traditional approaches. Additionally, we address challenges such as potential conflicts in tuning recommendations and the performance overhead associated with ML integration, providing insights for future research directions. This study is motivated by the need for autonomous tuning mechanisms to manage large-scale, heterogeneous workloads while answering key research questions, such as the following: (1) How can machine learning models be integrated into a DBMS to improve query optimization and workload management? (2) What performance improvements can be achieved through dynamic configuration tuning based on real-time workload patterns? Our results suggest that the proposed framework significantly reduces the need for manual database administration while effectively adapting to evolving workloads, offering a robust solution for modern large-scale data environments.

Sustainable energy efficient workflow classification and scheduling in geo distributed cloud datacenter

Data centers are a major source of carbon emissions and are subsequently contributing to global carbon footprints. Keeping in view of providing a sustainable solution to society, we have analyzed various factors that can help to achieve carbon neutrality and maximum sustainability. Our study pointed towards the need to follow a sustainable approach for incoming workflow throughout the life-cycle of Data centers. We analyzed that workloads need to be segregated before assigning them to the data centers so that energy-efficient resource allocation could be done. This paper demonstrates unsupervised learning techniques to cluster the incoming cloud workloads. The heterogeneous workloads were characterized using machine learning approaches and appropriate clusters were crafted. For analysis, Google Cluster Dataset is used. In order to improve the accuracy, data were normalized, and random samples of data were selected for clustering. The machine learning algorithms applied were able to successfully determine the appropriate clusters that can further be used for energy-efficient resource scheduling.

The Renoir Dataflow Platform: Efficient Data Processing without Complexity

Today, data analysis drives the decision-making process in virtually every human activity. This demands for software platforms that offer simple programming abstractions to express data analysis tasks and that can execute them in an efficient and scalable way. State-of-the-art solutions range from low-level programming primitives, which give control to the developer about communication and resource usage, but require significant effort to develop and optimize new algorithms, to high-level platforms that hide most of the complexities of parallel and distributed processing, but often at the cost of reduced efficiency.To reconcile these requirements, we developed Renoir, a novel distributed data processing platform written in Rust. Renoir provides a high-level dataflow programming model as mainstream data processing systems. It supports static and streaming data, it enables data transformations, grouping, aggregation, iterative computations, and time-based analytics, and it provides all these features incurring in a low overhead.In this paper, we present the programming model and the implementation details of Renoir. We evaluate it under heterogeneous workloads. We compare it with state-of-the-art solutions for data analysis and high-performance computing, as well as alternative research products, which offer different programming abstractions and implementation strategies. Renoir programs are compact and easy to write: developers need not care about low-level concerns such as resource usage, data serialization, concurrency control, and communication. At the same time, Renoir consistently presents comparable or better performance than competing solutions, by a large margin in several scenarios.We conclude that Renoir offers a good tradeoff between simplicity and performance, allowing developers to easily express complex data analysis tasks and achieve high performance and scalability.

Leveraging Dynamic and Heterogeneous Workload Knowledge to Boost the Performance of Index Advisors

Leveraging Dynamic and Heterogeneous Workload Knowledge to Boost the Performance of Index Advisors

Medley deep reinforcement learning-based workload offloading and cache placement decision in UAV-enabled MEC networks

Internet of Things devices generate a large number of heterogeneous workloads in real-time that require specific application to tackle, and the inability to communicate between devices and communication base stations due to complex scenarios is a thorny issue. Service caching play a key role in managing specific-request workload from devices, and unmanned aerial vehicles with computation and communication functions can effectively solve communication barrier between devices and ground base stations. In addition, the joint optimization of workload offloading and service cache placement is a key issue. Accordingly, we design an unmanned aerial vehicle-enabled mobile edge computing system with multiple devices, unmanned aerial vehicles and edge servers. The proposed framework takes into account the randomness of workload arrival, the time-varying nature of channel states, the limitations of the hosting service caching, and wireless communication blocking. Furthermore, we designed workload offloading and service caching hosting decision-making optimization problems to minimize the long-term weighted average latency and energy consumption costs. To tackle this joint optimization problem, we propose a request-specific workload offloading and service caching decision-making scheme based on the medley deep reinforcement learning scheme. To this end, the proposed scheme is decomposed into two-stage optimization subproblems: the workload offloading decision-making problem and the service caching hosting selection problem. In terms of the first subproblem, we model each device as a learning agent and propose the workloads offloading decision-making scheme based on multi-agent deep deterministic policy gradient. For the second subproblem, we present the decentralized double deep Q-learning scheme to tackle the service caching hosting policy. According to the comprehensive experimental results, the proposed scheme is able to converge rapidly on various parameter configurations and whose performance surpasses the other four baseline learning algorithms.

Multi-objective task scheduling method for cyber–physical–social systems in fog computing

Cyber–physical–social systems refer to new systems that involve a large number of Internet of Things (IoT) devices connected through an interconnected network. These systems gather data for various services such as transportation, energy, healthcare, etc. This data is then transmitted to the cloud layer for storage, processing, and access. To handle the significant workload offloaded from resource-constrained IoT devices without latency or bandwidth issues, a technology called fog computing has emerged as an intermediary layer between the cloud and IoT devices. The quality of service (QoS) provided by fog computing relies on efficient task scheduling to optimize energy consumption and maximum execution time. This task scheduling problem is known to be NP-hard and cannot be solved with high precision in a reasonable time using traditional techniques or existing metaheuristic-based task schedulers. Therefore, this study proposes a new multi-objective task scheduling approach based on a modified marine predators algorithm (MMPA) to simultaneously minimize energy consumption and make-span under the Pareto optimality theory. This variant is named the multi-objective MMPA (M2MPA). M2MPA enhances the performance of the MPA by incorporating the polynomial crossover operator to improve its exploration operator and the adaptive CF parameter to enhance its exploitation operator. The effectiveness of M2MPA is evaluated using eighteen tasks with different scales (small, large, and medium) and heterogeneous workloads assigned to two hundred fog devices with varying processing speeds. Finally, M2MPA is compared with six well-established techniques using various performance indicators, including carbon dioxide emission rate, flowtime, make-span, and energy consumption. The experimental findings demonstrate the substantial superiority of M2MPA over all competing algorithms across different performance indicators and the Wilcoxon rank-sum test. Quantitatively, M2MPA achieves an average make-span of 10.095, average energy consumption of 3.77E+06, average flowtime of 8.39E+03, and average carbon dioxide emission rate of 8.88E+06.

Placement of High Availability Geo-Distributed Data Centers in Emerging Economies

The data center markets in emerging economies are being built at a furious pace. When high availability is required, as it always is in the modern digital economy, the placement of geo-distributed data centers may be influenced by factors such as technician shortage and under-developed infrastructure, both of which are typical in emerging economies. Although the data center availability subject in general has been well studied, it remains unclear how rapid and unbalanced economic development in emerging economies may affect the availability of geo-distributed data centers and their cost of ownership. In this paper, we incorporate the unbalanced availability of infrastructure and technician into the data center placement. The problem is first formulated as a mixed integer nonlinear program (MINLP).To solve this potentially large scale problem, we transform it into a QCQP, capable of handling heterogeneous workloads. The resulting problem can then be efficiently solved by off-the-shelf optimization toolboxes. With real-life data in China, we show how unbalanced development of infrastructure and technician shortage may affect the placement of data centers, and analyze the tradeoff between cost and availability. Our results indicate that technician shortage and unbalanced network infrastructure will lead to increased cost and distinct data center placement strategies.

Resilience and load balancing in Fog networks: A Multi-Criteria Decision Analysis approach

The advent of Cloud Computing enabled the proliferation of Internet-of-Things (IoT) applications for smart environments. However, the distance of these resources makes them unsuitable for delay-sensitive applications. Hence, Fog Computing has emerged to provide such capabilities in proximity to end devices through distributed resources. These limited resources can collaborate to serve distributed IoT application workflows using the concept of stateless micro Fog service replicas, which provides resiliency and maintains service availability in the face of failures. Load balancing supports this collaboration by optimally assigning workloads to appropriate services, i.e., distributing the load among Fog nodes to fairly utilize compute and network resources and minimize execution delays. In this paper, we propose using ELECTRE, a Multi-Criteria Decision Analysis (MCDA) approach, to efficiently balance the load in Fog environments. We considered multiple objectives to make service selection decisions, including compute and network load information. We evaluate our approach in a realistic unbalanced topological setup with heterogeneous workload requirements. To the best of our knowledge, this is the first time ELECTRE-based methods are used to balance the load in Fog environments. Through simulations, we compared the performance of our proposed approach with traditional baseline methods that are commonly used in practice, namely random, Round-Robin, nearest node, and fastest service selection algorithms. In terms of the overall system performance, our approach outperforms these methods with up to 67% improvement.

Deep Learning-Based Job Placement in Distributed Machine Learning Clusters With Heterogeneous Workloads

Nowadays, most leading IT companies host a variety of distributed machine learning (ML) workloads in ML clusters to support AI-driven services, such as speech recognition, machine translation, and image processing. While multiple jobs are executed concurrently in a shared cluster to improve resource utilization, interference among co-located ML jobs can lead to significant performance downgrade. Existing cluster schedulers, such as YARN and Mesos, are interference-agnostic in their job placement, leading to suboptimal resource efficiency and usage. Some literature has studied interference-aware job placement policy, but relies on detailed workload profiling and interference modeling, which is not a general solution. In this work, we present Harmony, a deep learning-driven ML cluster scheduler that places heterogeneous training jobs (either with parameter server architecture or all-reduce architecture) in a manner that minimizes interference and maximizes performance (i.e., training completion time minimization). The design of Harmony is based on a carefully designed deep reinforcement learning (DRL) framework enhanced with reward modeling. The DRL integrates a dynamic sequence-to-sequence model with the state-of-the-art techniques to stabilize training and improve convergence, including actor-critic algorithm, job-aware action space exploration, multi-head attention, and experience replay. In view of a common lack of reward samples corresponding to different placement decisions, we build an auxiliary sequence-to-sequence reward prediction model, which is trained with historical samples and used for producing reward for unseen placement. Experiments using real ML workloads in a Kubernetes cluster of 6 GPU servers show that Harmony outperforms representative schedulers by 16%–42% in terms of average job completion time.

Modeling GPU Dynamic Parallelism for self similar density workloads

Dynamic Parallelism (DP) is a GPU programming abstraction that can make parallel computation more efficient for problems that exhibit heterogeneous workloads. With DP, GPU threads can launch kernels with more threads, recursively, producing a subdivision effect where resources are focused on the regions that exhibit more parallel work. Doing an optimal subdivision process is not trivial, as the combination of different parameters play a relevant role in the final performance of DP. Also, the current programming abstraction of DP relies on kernel recursion, which has performance overhead. This work presents a new subdivision cost model for problems that exhibit self similar density (SSD) workloads, useful for finding efficient subdivision schemes. Also, a new subdivision implementation free of recursion overhead is presented, named Adaptive Serial Kernels (ASK). Using the Mandelbrot set as a case study, the cost model shows that optimal performance is achieved when using {g∼32,r∼2,B∼32} for the initial subdivision, recurrent subdivision and stopping size, respectively. Experimental results agree with the theoretical parameters, confirming the usability of the cost model. In terms of performance, the ASK approach runs up to ∼60% faster than DP in the Mandelbrot set, and up to 12× faster than a basic exhaustive implementation, whereas DP is up to 7.5× faster. In terms of energy efficiency, ASK is up to ∼2× and ∼20× more energy efficient than DP and the exhaustive approach, respectively. These results put the subdivision cost model and the ASK approach as useful tools for analyzing the potential improvement of subdivision based approaches and for developing more efficient GPU-based libraries or fine-tune specific codes in research teams.

ML-Based Performance Modeling in SDN-Enabled Data Center Networks

Traffic optimization and smart buffering are fundamental to achieve both great application performance and resource efficiency in data centers with heterogeneous workloads, including incast and elephant traffics. However, general performance models providing insights on how various factors affect traffic performance metrics needed by these management functions are missing. For the special case of incast, the existing models are analytical ones, either tightly coupled with a particular protocol version or specific to certain empirical data. Motivated by this observation, this paper proposes an SDN-enabled machine-learning-based performance modeling approach in data center networks that leverages random forest predictions. Evaluations based on datasets constructed through intensive NS-3 simulations show that we can achieve accurate predictions of incast and elephant performance metrics based on various features. With this performance modeling capability, smart buffering schemes or traffic optimization algorithms could anticipate and efficiently optimize system parameters adjustment to achieve optimal performance continuously in data centers.

EEOA: Cost and Energy Efficient Task Scheduling in a Cloud-Fog Framework.

Cloud-fog computing is a wide range of service environments created to provide quick, flexible services to customers, and the phenomenal growth of the Internet of Things (IoT) has produced an immense amount of data on a daily basis. To complete tasks and meet service-level agreement (SLA) commitments, the provider assigns appropriate resources and employs scheduling techniques to efficiently manage the execution of received IoT tasks in fog or cloud systems. The effectiveness of cloud services is directly impacted by some other important criteria, such as energy usage and cost, which are not taken into account by many of the existing methodologies. To resolve the aforementioned problems, an effective scheduling algorithm is required to schedule the heterogeneous workload and enhance the quality of service (QoS). Therefore, a nature-inspired multi-objective task scheduling algorithm called the electric earthworm optimization algorithm (EEOA) is proposed in this paper for IoT requests in a cloud-fog framework. This method was created using the combination of the earthworm optimization algorithm (EOA) and the electric fish optimization algorithm (EFO) to improve EFO's potential to be exploited while looking for the best solution to the problem at hand. Concerning execution time, cost, makespan, and energy consumption, the suggested scheduling technique's performance was assessed using significant instances of real-world workloads such as CEA-CURIE and HPC2N. Based on simulation results, our proposed approach improves efficiency by 89%, energy consumption by 94%, and total cost by 87% over existing algorithms for the scenarios considered using different benchmarks. Detailed simulations demonstrate that the suggested approach provides a superior scheduling scheme with better results than the existing scheduling techniques.

Cloud-Native Workload Orchestration at the Edge: A Deployment Review and Future Directions.

Cloud-native computing principles such as virtualization and orchestration are key to transferring to the promising paradigm of edge computing. Challenges of containerization, operative models and scarce availability of established tools make a thorough review indispensable. Therefore, the authors have described the practical methods and tools found in the literature as well as in current community-led development projects, and have thoroughly exposed the future directions of the field. Container virtualization and its orchestration through Kubernetes have dominated the cloud computing domain, while major efforts have been recently recorded focused on the adaptation of these technologies to the edge. Such initiatives have addressed either the reduction of container engines and the development of specific tailored operating systems or the development of smaller K8s distributions and edge-focused adaptations (such as KubeEdge). Finally, new workload virtualization approaches, such as WebAssembly modules together with the joint orchestration of these heterogeneous workloads, seem to be the topics to pay attention to in the short to medium term.

A priority-aware scheduling framework for heterogeneous workloads in container-based cloud

A priority-aware scheduling framework for heterogeneous workloads in container-based cloud

Server temperature prediction using deep neural networks to assist thermal-aware scheduling

Thermal-Aware (TA) scheduling is a primitive thermal-management tool to avoid hotspots and attain a better thermal profile inside data centers. However, TA scheduling depends on the accuracy of server temperature calculation for making reliable scheduling decisions. Existing TA simulation frameworks rely on the CRAC, RC, and thermodynamics model to calculate the temperature of a server. However, these models are not very efficient computationally while calculating the temperature of computing nodes. Hence, there is a need for an efficient and lightweight temperature prediction model to fill this gap. Moreover, existing TA allocation and migration schemes neglect the workload heterogeneity regarding execution time (length) of user tasks in batch workloads. Ignoring task heterogeneity may lead to higher ambient temperature on its neighboring servers—assigning a hotter job with a longer duration to a server with higher ambient effect results in significantly higher cooling expenses.Our contribution in this paper is twofold: (1) firstly, we design and train a deep neural network to predict the temperature of servers; our proposed DNN model achieves 96.11% prediction accuracy, and (2) secondly, we propose a TA algorithm for job allocation and migration. Specifically, we consider the length and thermal profile of user tasks and servers during allocation and migration. We compare the proposed strategy against existing TA scheduling designs such as TA Scheduling Algorithm (TASA) and TA Control Strategy (TACS) using simulations. Results demonstrate that the proposed TA approach significantly reduces the overall energy consumption by up to 12.03% and 8.28% compared to the TASA and TACS, respectively.

A Case for Fine-grain Coherence Specialization in Heterogeneous Systems

Hardware specialization is becoming a key enabler of energy-efficient performance. Future systems will be increasingly heterogeneous, integrating multiple specialized and programmable accelerators, each with different memory demands. Traditionally, communication between accelerators has been inefficient, typically orchestrated through explicit DMA transfers between different address spaces. More recently, industry has proposed unified coherent memory which enables implicit data movement and more data reuse, but often these interfaces limit the coherence flexibility available to heterogeneous systems.This paper demonstrates the benefits of fine-grained coherence specialization for heterogeneous systems. We propose an architecture that enables low-complexity independent specialization of each individual coherence request in heterogeneous workloads by building upon a simple and flexible baseline coherence interface, Spandex. We then describe how to optimize individual memory requests to improve cache reuse and performance-critical memory latency in emerging heterogeneous workloads. Collectively, our techniques enable significant gains, reducing execution time by up to 61% or network traffic by up to 99% while adding minimal complexity to the Spandex protocol.

Energy-Aware Heuristic Scheduling Using Bin Packing MapReduce Scheduler for Heterogeneous Workloads Performance in Big Data

Energy-Aware Heuristic Scheduling Using Bin Packing MapReduce Scheduler for Heterogeneous Workloads Performance in Big Data

Understanding Performance of a Vulnerable Heterogeneous Edge Data Center: A Modeling Approach

Abstract Internet of Things (IoT) jobs not only require computational resources but also are delay-sensitive and security-sensitive. Edge computing emerges as a promising paradigm to improve the quality of experience for IoT users. Edge computing faces many security threats, perhaps even more than traditional data centers. With a growing amount of data offloaded to Edge Data Centers (EDCs), the EDC performance needs to be considered and evaluated carefully for improving the vulnerable EDC resource utilization while satisfying IoT job requirements. This paper develops an analytical model, which can capture the dynamics of an EDC system with the following features: (i) The system is under heterogeneous workloads; (ii) the system is subject to attacks, which prevent equipment units in the system from providing service and (iii) the jobs in the system are delay-sensitive. Namely, the job processing fails before the processing is completed. Based on the proposed model, we develop formulas for performance and profit metrics and conduct a series of simulation experiments to verify the correctness and accuracy of our model. Finally, through our model, we evaluate the performance of the EDC, and we offer solutions for EDC administrators to maximize profit.

A Resource Recommendation Model for Heterogeneous Workloads in Fog-Based Smart Factory Environment

The wide deployment of advanced robots with industrial IoT (IIoT) technologies in smart factories generates a large volume of data during production and a wide variety of data processing workloads are launched to maintain productivity and safety of smart manufacture. The emerging fog computing paradigm offers a promising solution to enhancing data processing performance in a smart factory environment while on the other hand brings in new challenges to resource management, which call for a more effective approach for recommending resource configurations to heterogeneous workloads. In this paper, we propose an Optimized Recommendations of Heterogeneous Resource Configurations (ORHRC) model that employs machine learning techniques to provide resource configuration recommendations for the heterogeneous workloads in a fog computing-based smart factory environment. ORHRC learns a recommendation model by leveraging the operating characteristics and execution time of workloads on fog servers with different configurations. We also design a decision model in ORHRC to further improve prediction accuracy and reduce operational overheads. Experiment results show that ORHRC outperforms the state of art configuration recommendation methods in terms of average prediction accuracy. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —The various data processing workloads in a smart factory environment need to be processed by the computational resources with optimal configurations for meeting their performance requirements. In this paper, we employ machine learning technologies for enabling automatic recommendation of resource configurations to heterogeneous workloads. Specifically, we develop an Optimized Recommendations of Heterogeneous Resource Configurations (ORHRC) model that can identify the optimal resource configurations for various workloads. We also conducted extensive experiments that verify the effectiveness of the proposed ORHRC model.

A Heterogeneous In-Memory Computing Cluster for Flexible End-to-End Inference of Real-World Deep Neural Networks

Deployment of modern TinyML tasks on small battery-constrained IoT devices requires high computational energy efficiency. Analog In-Memory Computing (IMC) using non-volatile memory (NVM) promises major efficiency improvements in deep neural network (DNN) inference and serves as on-chip memory storage for DNN weights. However, IMC’s functional flexibility limitations and their impact on performance, energy, and area efficiency are not yet fully understood at the system level. To target practical end-to-end IoT applications, IMC arrays must be enclosed in heterogeneous programmable systems, introducing new system-level challenges which we aim at addressing in this work. We present a heterogeneous tightly-coupled clustered architecture integrating 8 RISC-V cores, an in-memory computing accelerator (IMA), and digital accelerators. We benchmark the system on a highly heterogeneous workload such as the Bottleneck layer from a MobileNetV2, showing <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$11.5\times $ </tex-math></inline-formula> performance and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$9.5\times $ </tex-math></inline-formula> energy efficiency improvements, compared to highly optimized parallel execution on the cores. Furthermore, we explore the requirements for end-to-end inference of a full mobile-grade DNN (MobileNetV2) in terms of IMC array resources, by scaling up our heterogeneous architecture to a multi-array accelerator. Our results show that our solution, on the end-to-end inference of the MobileNetV2, is one order of magnitude better in terms of execution latency than existing programmable architectures and two orders of magnitude better than state-of-the-art heterogeneous solutions integrating in-memory computing analog cores.