arXiv subjects

Lauritz Thamsen

Publications and source records attributed to Lauritz Thamsen.

At least 37 records · Page 2Linked to original sources

arXiv2025

Flora: Efficient Cloud Resource Selection for Big Data Processing via Job Classification

Distributed dataflow systems like Spark and Flink enable data-parallel processing of large datasets on clusters of cloud resources. Yet, selecting appropriate computational resources for dataflow jobs is often challenging. For efficient execution, individual resource allocations, such as memory and CPU cores, must meet the specific resource demands of the job. Meanwhile, the choices of cloud configurations are often plentiful, especially in public clouds, and the current cost of the available resource options can fluctuate. Addressing this challenge, we present Flora, a low-overhead approach to cost-optimizing cloud cluster configurations for big data processing. Flora lets users categorize jobs according to their data access patterns and derives suitable cluster resource configurations from executions of test jobs of the same category, considering current resource costs. In our evaluation on a new dataset comprising 180 Spark job executions on Google Cloud, Flora's cluster resource selections exhibit an average deviation below 6% from the most cost-optimal solution, with a maximum deviation below 24%.

cs.DC↗

arXiv2025

Sizey: Memory-Efficient Execution of Scientific Workflow Tasks

As the amount of available data continues to grow in fields as diverse as bioinformatics, physics, and remote sensing, the importance of scientific workflows in the design and implementation of reproducible data analysis pipelines increases. When developing workflows, resource requirements must be defined for each type of task in the workflow. Typically, task types vary widely in their computational demands because they are simply wrappers for arbitrary black-box analysis tools. Furthermore, the resource consumption for the same task type can vary considerably as well due to different inputs. Since underestimating memory resources leads to bottlenecks and task failures, workflow developers tend to overestimate memory resources. However, overprovisioning of memory wastes resources and limits cluster throughput. Addressing this problem, we propose Sizey, a novel online memory prediction method for workflow tasks. During workflow execution, Sizey simultaneously trains multiple machine learning models and then dynamically selects the best model for each workflow task. To evaluate the quality of the model, we introduce a novel resource allocation quality (RAQ) score based on memory prediction accuracy and efficiency. Sizey's prediction models are retrained and re-evaluated online during workflow execution, continuously incorporating metrics from completed tasks. Our evaluation with a prototype implementation of Sizey uses metrics from six real-world scientific workflows from the popular nf-core framework and shows a median reduction in memory waste over time of 24.68% compared to the respective best-performing state-of-the-art baseline.

cs.DC↗

arXiv2025

Experimentally Evaluating the Resource Efficiency of Big Data Autoscaling

Distributed dataflow systems like Spark and Flink enable data-parallel processing of large datasets on clusters. Yet, selecting appropriate computational resources for dataflow jobs is often challenging. For efficient execution, individual resource allocations, such as memory and CPU cores, must meet the specific resource requirements of the job. An alternative to selecting a static resource allocation for a job execution is autoscaling as implemented for example by Spark. In this paper, we evaluate the resource efficiency of autoscaling batch data processing jobs based on resource demand both conceptually and experimentally by analyzing a new dataset of Spark job executions on Google Dataproc Serverless. In our experimental evaluation, we show that there is no significant resource efficiency gain over static resource allocations. We found that the inherent conceptual limitations of such autoscaling approaches are the inelasticity of node size as well as the inelasticity of the ratio of memory to CPU cores.

cs.DC↗

arXiv2024

Ponder: Online Prediction of Task Memory Requirements for Scientific Workflows

Scientific workflows are used to analyze large amounts of data. These workflows comprise numerous tasks, many of which are executed repeatedly, running the same custom program on different inputs. Users specify resource allocations for each task, which must be sufficient for all inputs to prevent task failures. As a result, task memory allocations tend to be overly conservative, wasting precious cluster resources, limiting overall parallelism, and increasing workflow makespan. In this paper, we first benchmark a state-of-the-art method on four real-life workflows from the nf-core workflow repository. This analysis reveals that certain assumptions underlying current prediction methods, which typically were evaluated only on simulated workflows, cannot generally be confirmed for real workflows and executions. We then present Ponder, a new online task-sizing strategy that considers and chooses between different methods to cater to different memory demand patterns. We implemented Ponder for Nextflow and made the code publicly available. In an experimental evaluation that also considers the impact of memory predictions on scheduling, Ponder improves Memory Allocation Quality on average by 71.0% and makespan by 21.8% in comparison to a state-of-the-art method. Moreover, Ponder produces 93.8% fewer task failures.

cs.DC↗

arXiv2024

KS+: Predicting Workflow Task Memory Usage Over Time

Scientific workflow management systems enable the reproducible execution of data analysis pipelines on cluster infrastructures managed by resource managers such as Kubernetes, Slurm, or HTCondor. These resource managers require resource estimates for each workflow task to be executed on one of the cluster nodes. However, task resource consumption varies significantly between different tasks and for the same task with different inputs. Furthermore, resource consumption also fluctuates during a task's execution. As a result, manually configuring static memory allocations is error-prone, often leading users to overestimate memory usage to avoid costly failures from under-provisioning, which results in significant memory wastage. We propose KS+, a method that predicts a task's memory consumption over time depending on its inputs. For this, KS+ dynamically segments the task execution and predicts the memory required for each segment. Our experimental evaluation shows an average reduction in memory wastage of 38% compared to the best-performing state-of-the-art baseline for two real-world workflows from the popular nf-core repository.

cs.DC↗

arXiv2024

Predicting Dynamic Memory Requirements for Scientific Workflow Tasks

With the increasing amount of data available to scientists in disciplines as diverse as bioinformatics, physics, and remote sensing, scientific workflow systems are becoming increasingly important for composing and executing scalable data analysis pipelines. When writing such workflows, users need to specify the resources to be reserved for tasks so that sufficient resources are allocated on the target cluster infrastructure. Crucially, underestimating a task's memory requirements can result in task failures. Therefore, users often resort to overprovisioning, resulting in significant resource wastage and decreased throughput. In this paper, we propose a novel online method that uses monitoring time series data to predict task memory usage in order to reduce the memory wastage of scientific workflow tasks. Our method predicts a task's runtime, divides it into k equally-sized segments, and learns the peak memory value for each segment depending on the total file input size. We evaluate the prototype implementation of our method using workflows from the publicly available nf-core repository, showing an average memory wastage reduction of 29.48% compared to the best state-of-the-art approach.

cs.DC↗

arXiv2024

Privacy-Preserving Sharing of Data Analytics Runtime Metrics for Performance Modeling

Performance modeling for large-scale data analytics workloads can improve the efficiency of cluster resource allocations and job scheduling. However, the performance of these workloads is influenced by numerous factors, such as job inputs and the assigned cluster resources. As a result, performance models require significant amounts of training data. This data can be obtained by exchanging runtime metrics between collaborating organizations. Yet, not all organizations may be inclined to publicly disclose such metadata. We present a privacy-preserving approach for sharing runtime metrics based on differential privacy and data synthesis. Our evaluation on performance data from 736 Spark job executions indicates that fully anonymized training data largely maintains performance prediction accuracy, particularly when there is minimal original data available. With 30 or fewer available original data samples, the use of synthetic training data resulted only in a one percent reduction in performance model accuracy on average.

cs.DC↗

arXiv2024

Demeter: Resource-Efficient Distributed Stream Processing under Dynamic Loads with Multi-Configuration Optimization

Distributed Stream Processing (DSP) focuses on the near real-time processing of large streams of unbounded data. To increase processing capacities, DSP systems are able to dynamically scale across a cluster of commodity nodes, ensuring a good Quality of Service despite variable workloads. However, selecting scaleout configurations which maximize resource utilization remains a challenge. This is especially true in environments where workloads change over time and node failures are all but inevitable. Furthermore, configuration parameters such as memory allocation and checkpointing intervals impact performance and resource usage as well. Sub-optimal configurations easily lead to high operational costs, poor performance, or unacceptable loss of service. In this paper, we present Demeter, a method for dynamically optimizing key DSP system configuration parameters for resource efficiency. Demeter uses Time Series Forecasting to predict future workloads and Multi-Objective Bayesian Optimization to model runtime behaviors in relation to parameter settings and workload rates. Together, these techniques allow us to determine whether or not enough is known about the predicted workload rate to proactively initiate short-lived parallel profiling runs for data gathering. Once trained, the models guide the adjustment of multiple, potentially dependent system configuration parameters ensuring optimized performance and resource usage in response to changing workload rates. Our experiments on a commodity cluster using Apache Flink demonstrate that Demeter significantly improves the operational efficiency of long-running benchmark jobs.

cs.DC↗

arXiv2024

Towards a Peer-to-Peer Data Distribution Layer for Efficient and Collaborative Resource Optimization of Distributed Dataflow Applications

Performance modeling can help to improve the resource efficiency of clusters and distributed dataflow applications, yet the available modeling data is often limited. Collaborative approaches to performance modeling, characterized by the sharing of performance data or models, have been shown to improve resource efficiency, but there has been little focus on actual data sharing strategies and implementation in production environments. This missing building block holds back the realization of proposed collaborative solutions. In this paper, we envision, design, and evaluate a peer-to-peer performance data sharing approach for collaborative performance modeling of distributed dataflow applications. Our proposed data distribution layer enables access to performance data in a decentralized manner, thereby facilitating collaborative modeling approaches and allowing for improved prediction capabilities and hence increased resource efficiency. In our evaluation, we assess our approach with regard to deployment, data replication, and data validation, through experiments with a prototype implementation and simulation, demonstrating feasibility and allowing discussion of potential limitations and next steps.

cs.DC↗

arXiv2024

FedZero: Leveraging Renewable Excess Energy in Federated Learning

Federated Learning (FL) is an emerging machine learning technique that enables distributed model training across data silos or edge devices without data sharing. Yet, FL inevitably introduces inefficiencies compared to centralized model training, which will further increase the already high energy usage and associated carbon emissions of machine learning in the future. One idea to reduce FL's carbon footprint is to schedule training jobs based on the availability of renewable excess energy that can occur at certain times and places in the grid. However, in the presence of such volatile and unreliable resources, existing FL schedulers cannot always ensure fast, efficient, and fair training. We propose FedZero, an FL system that operates exclusively on renewable excess energy and spare capacity of compute infrastructure to effectively reduce a training's operational carbon emissions to zero. Using energy and load forecasts, FedZero leverages the spatio-temporal availability of excess resources by selecting clients for fast convergence and fair participation. Our evaluation, based on real solar and load traces, shows that FedZero converges significantly faster than existing approaches under the mentioned constraints while consuming less energy. Furthermore, it is robust to forecasting errors and scalable to tens of thousands of clients.

cs.LG↗

arXiv2023

The Common Workflow Scheduler Interface: Status Quo and Future Plans

Nowadays, many scientific workflows from different domains, such as Remote Sensing, Astronomy, and Bioinformatics, are executed on large computing infrastructures managed by resource managers. Scientific workflow management systems (SWMS) support the workflow execution and communicate with the infrastructures' resource managers. However, the communication between SWMS and resource managers is complicated by a) inconsistent interfaces between SMWS and resource managers and b) the lack of support for workflow dependencies and workflow-specific properties. To tackle these issues, we developed the Common Workflow Scheduler Interface (CWSI), a simple yet powerful interface to exchange workflow-related information between a SWMS and a resource manager, making the resource manager workflow-aware. The first prototype implementations show that the CWSI can reduce the makespan already with simple but workflow-aware strategies up to 25%. In this paper, we show how existing workflow resource management research can be integrated into the CWSI.

cs.DC↗

arXiv2023

Karasu: A Collaborative Approach to Efficient Cluster Configuration for Big Data Analytics

Selecting the right resources for big data analytics jobs is hard because of the wide variety of configuration options like machine type and cluster size. As poor choices can have a significant impact on resource efficiency, cost, and energy usage, automated approaches are gaining popularity. Most existing methods rely on profiling recurring workloads to find near-optimal solutions over time. Due to the cold-start problem, this often leads to lengthy and costly profiling phases. However, big data analytics jobs across users can share many common properties: they often operate on similar infrastructure, using similar algorithms implemented in similar frameworks. The potential in sharing aggregated profiling runs to collaboratively address the cold start problem is largely unexplored. We present Karasu, an approach to more efficient resource configuration profiling that promotes data sharing among users working with similar infrastructures, frameworks, algorithms, or datasets. Karasu trains lightweight performance models using aggregated runtime information of collaborators and combines them into an ensemble method to exploit inherent knowledge of the configuration search space. Moreover, Karasu allows the optimization of multiple objectives simultaneously. Our evaluation is based on performance data from diverse workload executions in a public cloud environment. We show that Karasu is able to significantly boost existing methods in terms of performance, search time, and cost, even when few comparable profiling runs are available that share only partial common characteristics with the target job.

cs.DC↗

arXiv2023

Lotaru: Locally Predicting Workflow Task Runtimes for Resource Management on Heterogeneous Infrastructures

Many resource management techniques for task scheduling, energy and carbon efficiency, and cost optimization in workflows rely on a-priori task runtime knowledge. Building runtime prediction models on historical data is often not feasible in practice as workflows, their input data, and the cluster infrastructure change. Online methods, on the other hand, which estimate task runtimes on specific machines while the workflow is running, have to cope with a lack of measurements during start-up. Frequently, scientific workflows are executed on heterogeneous infrastructures consisting of machines with different CPU, I/O, and memory configurations, further complicating predicting runtimes due to different task runtimes on different machine types. This paper presents Lotaru, a method for locally predicting the runtimes of scientific workflow tasks before they are executed on heterogeneous compute clusters. Crucially, our approach does not rely on historical data and copes with a lack of training data during the start-up. To this end, we use microbenchmarks, reduce the input data to quickly profile the workflow locally, and predict a task's runtime with a Bayesian linear regression based on the gathered data points from the local workflow execution and the microbenchmarks. Due to its Bayesian approach, Lotaru provides uncertainty estimates that can be used for advanced scheduling methods on distributed cluster infrastructures. In our evaluation with five real-world scientific workflows, our method outperforms two state-of-the-art runtime prediction baselines and decreases the absolute prediction error by more than 12.5%. In a second set of experiments, the prediction performance of our method, using the predicted runtimes for state-of-the-art scheduling, carbon reduction, and cost prediction, enables results close to those achieved with perfect prior knowledge of runtimes.

cs.DC↗

arXiv2023

How Workflow Engines Should Talk to Resource Managers: A Proposal for a Common Workflow Scheduling Interface

Scientific workflow management systems (SWMSs) and resource managers together ensure that tasks are scheduled on provisioned resources so that all dependencies are obeyed, and some optimization goal, such as makespan minimization, is achieved. In practice, however, there is no clear separation of scheduling responsibilities between an SWMS and a resource manager because there exists no agreed-upon separation of concerns between their different components. This has two consequences. First, the lack of a standardized API to exchange scheduling information between SWMSs and resource managers hinders portability. It incurs costly adaptations when a component should be replaced by a different one (e.g., an SWMS with another SWMS on the same resource manager). Second, due to overlapping functionalities, current installations often actually have two schedulers, both making partial scheduling decisions under incomplete information, leading to suboptimal workflow scheduling. In this paper, we propose a simple REST interface between SWMSs and resource managers, which allows any SWMS to pass dynamic workflow information to a resource manager, enabling maximally informed scheduling decisions. We provide an implementation of this API as an example, using Nextflow as an SWMS and Kubernetes as a resource manager. Our experiments with nine real-world workflows show that this strategy reduces makespan by up to 25.1% and 10.8% on average compared to the standard Nextflow/Kubernetes configuration. Furthermore, a more widespread implementation of this API would enable leaner code bases, a simpler exchange of components of workflow systems, and a unified place to implement new scheduling algorithms.

cs.DC↗

arXiv2023

Selecting Efficient Cluster Resources for Data Analytics: When and How to Allocate for In-Memory Processing?

Distributed dataflow systems such as Apache Spark or Apache Flink enable parallel, in-memory data processing on large clusters of commodity hardware. Consequently, the appropriate amount of memory to allocate to the cluster is a crucial consideration. In this paper, we analyze the challenge of efficient resource allocation for distributed data processing, focusing on memory. We emphasize that in-memory processing with in-memory data processing frameworks can undermine resource efficiency. Based on the findings of our trace data analysis, we compile requirements towards an automated solution for efficient cluster resource allocation.

cs.DC↗

arXiv2023

Towards a Real-Time IoT: Approaches for Incoming Packet Processing in Cyber-Physical Systems

Embedded real-time devices for monitoring, controlling, and collaboration purposes in cyber-physical systems are now commonly equipped with IP networking capabilities. However, the reception and processing of IP packets generates workloads in unpredictable frequencies as networks are outside of a developer's control and difficult to anticipate, especially when networks are connected to the internet. As of now, embedded network controllers and IP stacks are not designed for real-time capabilities, even when used in real-time environments and operating systems. Our work focuses on real-time aware packet reception from open network connections, without a real-time networking infrastructure. This article presents two experimentally evaluated modifications to the IP processing subsystem and embedded network interface controllers of constrained IoT devices. The first, our software approach, introduces early packet classification and priority-aware processing in the network driver. In our experiments this allowed the network subsystem to remain active at a seven-fold increase in network traffic load before disabling the receive interrupts as a last resort. The second, our hardware approach, makes changes to the network interface controller, applying interrupt moderation based on real-time priorities to minimize the number of network-generated interrupts. Furthermore, this article provides an outlook on how the software and hardware approaches can be combined in a co-designed packet receive architecture.

cs.NI↗

arXiv2023

Towards Energy Consumption and Carbon Footprint Testing for AI-driven IoT Services

Energy consumption and carbon emissions are expected to be crucial factors for Internet of Things (IoT) applications. Both the scale and the geo-distribution keep increasing, while Artificial Intelligence (AI) further penetrates the "edge" in order to satisfy the need for highly-responsive and intelligent services. To date, several edge/fog emulators are catering for IoT testing by supporting the deployment and execution of AI-driven IoT services in consolidated test environments. These tools enable the configuration of infrastructures so that they closely resemble edge devices and IoT networks. However, energy consumption and carbon emissions estimations during the testing of AI services are still missing from the current state of IoT testing suites. This study highlights important questions that developers of AI-driven IoT services are in need of answers, along with a set of observations and challenges, aiming to help researchers designing IoT testing and benchmarking suites to cater to user needs.

cs.DC↗

arXiv2023

Ruya: Memory-Aware Iterative Optimization of Cluster Configurations for Big Data Processing

Selecting appropriate computational resources for data processing jobs on large clusters is difficult, even for expert users like data engineers. Inadequate choices can result in vastly increased costs, without significantly improving performance. One crucial aspect of selecting an efficient resource configuration is avoiding memory bottlenecks. By knowing the required memory of a job in advance, the search space for an optimal resource configuration can be greatly reduced. Therefore, we present Ruya, a method for memory-aware optimization of data processing cluster configurations based on iteratively exploring a narrowed-down search space. First, we perform job profiling runs with small samples of the dataset on just a single machine to model the job's memory usage patterns. Second, we prioritize cluster configurations with a suitable amount of total memory and within this reduced search space, we iteratively search for the best cluster configuration with Bayesian optimization. This search process stops once it converges on a configuration that is believed to be optimal for the given job. In our evaluation on a dataset with 1031 Spark and Hadoop jobs, we see a reduction of search iterations to find an optimal configuration by around half, compared to the baseline.

cs.DC↗