Distributed and Parallel Systems

,

Paper accepted: DFARM: A deadline-aware fault-tolerant scheduler for cloud computing

Cluster Computing

DFARM: A deadline-aware fault-tolerant scheduler for cloud computing

Authors: Ahmad Awan, Muhammad Aleem, Altaf Hussain, Radu Prodan

Abstract:

Cloud computing has become popular for small businesses due to its cost-effectiveness and the ability to acquire necessary on-demand services, including software, hardware, network, etc., anytime around the globe. Efficient job scheduling in the Cloud is essential to optimize operational costs in data centers. Therefore, scheduling should consider assigning tasks to Virtual Machines (VMs) in a Cloud environment in such a manner that could speed up execution, maximize resource utilization, and meet users’ SLA and other constraints such as deadlines. For this purpose, the tasks can be prioritized based on their deadlines and task lengths, and the resources could be provisioned and released as needed. Moreover, to cope with unexpected execution situations or hardware failures, a fault-tolerance mechanism could be employed based on hybrid replication and the re-submission method. Most of the existing techniques tend to improve performance. However, their pitfall lies in certain aspects such as either those techniques prioritize tasks based on a singular value (e.g., usually deadline), only utilize a singular fault tolerance mechanism, or try to release resources that cause more overhead immediately. This research work proposes a new scheduler called the Deadline and fault-aware task Adjusting and Resource Managing (DFARM) scheduler, the scheduler dynamically acquires resources and schedules deadline-constrained tasks by considering both their length and deadlines while providing fault tolerance through the hybrid replication-resubmission method. Besides acquiring resources, it also releases resources based on their boot time to lessen costs due to reboots. The performance of the DFARM scheduler is compared to other scheduling algorithms, such as Random Selection, Round Robin, Minimum Completion Time, RALBA, and OG-RADL. With a comparable execution performance, the proposed DFARM scheduler reduces task-rejection rates by $2.34 – 9.53$ times compared to the state-of-the-art schedulers using two benchmark datasets.

,

Paper accepted: Evaluation of Storage Placement in Computing Continuum for a Robotic Application: A simulation-based performance analysis

Journal of Grid Computing

Authors: Zeinab Bakhshi, Guillermo Rodriguez-Navas, Hans Hansson, Radu Prodan

Abstract:

This paper analyzes a persistent storage method’s timing performance for distributed container-based architectures in industrial control applications. The method focuses on ensuring data availability and consistency while accommodating faults. The analysis considers four aspects: placement strategy, design options, data size, and evaluation under faulty conditions. Experimental results considering the timing constraints in industrial applications indicate that the storage solution can meet critical deadlines, particularly under specific failure patterns. Moreover, the method is applicable for evaluating timing constraints in other container-based critical applications that require persistent storage.Further comparison results reveal that, while the method may underperform current centralized solutions under fault-free conditions, it outperforms the centralized solutions in failure scenarios.

,

Sahar Nasirihaghighi on Event Recognition with Hybrid Transformers at MMM 2024

On February 1st, 2024, Sahar Nasirihaghighi presented our work on ‘Event Recognition in Laparoscopic Gynecology Videos with Hybrid Transformers’ at this year’s International Conference on Multimedia Modeling (MMM 2024) in Amsterdam, The Netherlands.

Authors: Sahar Nasirihaghighi, Negin Ghamsarian, Heinrich Husslein, Klaus Schoeffmann

Abstract: Analyzing laparoscopic surgery videos presents a complex and multifaceted challenge, with applications including surgical training, intra-operative surgical complication prediction, and post-operative surgical assessment. Identifying crucial events within these videos is a significant prerequisite in a majority of these applications. In this paper, we introduce a comprehensive dataset tailored for relevant event recognition in laparoscopic gynecology videos. Our dataset includes annotations for critical events associated with major intra-operative challenges and post-operative complications. To validate the precision of our annotations, we assess event recognition performance using several CNN-RNN architectures. Furthermore, we introduce and evaluate a hybrid transformer architecture coupled with a customized training-inference framework to recognize four specific events in laparoscopic surgery videos. Leveraging the Transformer networks, our proposed architecture harnesses inter-frame dependencies to counteract the adverse effects of relevant content occlusion, motion blur, and surgical scene variation, thus significantly enhancing event recognition accuracy. Moreover, we present a frame sampling strategy designed to manage variations in surgical scenes and the surgeons’ skill level, resulting in event recognition with high temporal resolution. We empirically demonstrate the superiority of our proposed methodology in event recognition compared to conventional CNN-RNN architectures through a series of extensive experiments.

 

,

Radu Prodan at COST Action CERCIRAS Joint Meeting in Montpellier, France

An EU funding programme enabling researchers to set up their own interdisciplinary research networks in Europe and beyond. #COSTactions

Representing Ireland with Prof. Horacio González-Vélez of National College of Ireland at the partner meeting of the Cost Action Cerciras – Connecting Education and Research Communities for an Innovative Resource Aware Society in Montpellier today.

 

 

 

 

 

 

 

 

 

 


Great alignment with several EU skills projects like ARISA – AI Skills, ESSA Software Skills Digital4Business and Digital4Security by facilitating transversal insights.

,

ITEC team participated at the the 16th IEEE/ACM International Conference on Utility and Cloud Computing

In a noteworthy presence at the 16th IEEE/ACM International Conference on Utility and Cloud Computing (UCC 2023), Dragi Kimovski and Narges Mehran presented three workshop papers:

  1. Marcus Reitzl and Dragi Kimovski, “Multi-Objective Optimisation of Container Orchestration Systems”,
  2. Narges Mehran, Dragi Kimovski, Hermann Hellwagner, Dumitru Roman, Ahmet Soylu, and Radu Prodan, “Scheduling on the Computing Continuum: A Survey “,
  3. Narges Mehran, Arman Haghighi, Pedram Aminharati, Nikolay Nikolov, Ahmet Soylu, Dimitru Roman and Radu Prodan, “Comparison of Microservice Call Rate Prediction for Replication in the Cloud”. 

 Additionally, Dragi Kimovski took on the role of a session chair, leading discussions on the intricacies of scheduling in the computing continuum. 

,

COST Action Training School in summer 2024 @ ITEC

Radu Prodan/ITEC will organise the CERCIRAS Training School (on behalf of COST Action CA19135) from August 26 – 30, 2024 at the University of Klagenfurt. Details and programs will be available soon.

Radu Prodan @ PhD defense of Ruyue Xin, University of Amsterdam

Radu Prodan participated on November 1, 2023, as an external opponent at the PhD defense of Ruyue Xin ((Title of the dissertation: Towards Effective Performance Diagnosis for Distributed Applications), supervised by Dr. Zhiming Zhao, Dr. Paola Grosso and Prof. Cees de Laat at the University of Amsterdam, Netherlands.

 

,

Graph Massivizer Project at EBDVF23 in Valencia, Spain

The Graph Massivizer Project is part of the European Big Data Value Forum!

The team participates in a session exploring the latest in #KnowledgeGraph technology with real-world use cases in agrifood, industry 4.0 and healthcare. As part of the session, metaphacts GmbH founder Peter Haase will discuss the importance of knowledge graphs as a foundational layer for #AI applications.

,

Mastering Dark Data: Crafting Big Data Pipelines from Event Logs using the DataCloud Toolbox

The @DataCloud2020 dissemination Workshop, oriented by project partners @SINTEF, took place on the 26th of October as part of the @icpm_conf 2023, hosted by @SapienzaRoma. Narges Mehran participated for UNI-KLU.
@DataCloud2020 booth stand at the Auditorium Antonianum for the 5th International Conference on Process Mining (@icpm_conf 2023).

Radu Prodan talked about “Cloud Computing in Klagenfurt” in Wagenigen, Netherlands

The Symposium “The Data Science and Artificial Intelligence (DSAI) carnival” took place on the 19th of October 2023 at the Wageningen University & Research Campus and was organized in collaboration with the Wageningen Data Competence Center (WDCC).

This symposium provided an in-depth examination of cutting-edge themes from areas such as the Web, Semantic Web, linked data and knowledge graphs, LLMs, MLOps, cloud computing, data infrastructures and data space, FAIR data management, and related developments.

Leading experts shared the latest research and applications in these areas, fostering collaboration and offering insights into emerging trends.

The event concluded with the inaugural lecture of Prof. Dr. Anna Fensel.