,

Paper accepted: Energy-Quality-aware Variable Framerate Pareto-Front for Adaptive Video Streaming

Authors: Prajit T Rajendran (Universite Paris-Saclay), Samira Afzal (Alpen-Adria-Universität Klagenfurt), Vignesh V Menon (Fraunhofer HHI), Christian Timmerer (Alpen-Adria-Universität Klagenfurt)

Conference: IEEE Visual Communications and Image Processing (IEEE VCIP 2024)

Abstract: Optimizing framerate for a given bitrate-spatial resolution pair in adaptive video streaming is essential to maintain perceptual quality while considering decoding complexity. Low framerates at low bitrates reduce compression artifacts and decrease decoding energy. We propose a novel method, Decoding-complexity aware Framerate Prediction (DECODRA), which employs a Variable Framerate Pareto-front approach to predict an optimized framerate that minimizes decoding energy under quality degradation constraints. DECODRA dynamically adjusts the framerate based on current bitrate and spatial resolution, balancing trade-offs between framerate, perceptual quality, and decoding complexity. Extensive experimentation with the Inter-4K dataset demonstrates DECODRA’s effectiveness, yielding an average PSNR and VMAF increase of 0.87 dB and 5.14 points, respectively, for the same bitrate compared to the default 60 fps encoding. Additionally, DECODRA achieves an average reduction in decoding energy consumption of 13.27 %, enhancing the viewing experience, extending mobile device battery life, and reducing the energy footprint of streaming services.

,

Paper accepted – StoreLess: Serverless workflow scheduling with federated storage in sky computing

Authors: Sashko Ristov, Mika Hautz, Philipp Gritsch, Stefan Nastic, Radu Prodan, Michael Felderer

ICSOC 2024: 22nd International Conference on Service-Oriented Computing https://icsoc2024.redcad.tn/

Abstract:

We observe irregular data transfer performance across federated serverless infrastructures (sometimes faster across providers than colocated), making the entire workflow scheduling even more challenging in federated FaaS and sky computing. This paper introduces STORELESS – a novel workflow scheduler and heuristic algorithm for serverless storage attachments that dynamically selects, provisions, and configures suitable function deployments and storage backends from the federated serverless infrastructure. STORELESS improves workflow execution time by up to 30% by running cross-regional setup compared to the state-of-the-art.

 

,

Haleh @ First International Workshop on Scaling Knowledge Graphs for Industry” in Conference “Semantics 24: 20th International Conference on Semantic Systems”

Haleh gave a presentation at the first International Workshop on Scaling Knowledge Graphs for Industry (co-located with the 20th International Conference on Semantic Systems – SEMANTICS 2024)  – in Amsterdam, Sept. 17-19, 2024.

Title: “Modeling and Generating Extreme Volumes of Financial Synthetic Time-Series Data with Knowledge Graphs”

Authors: Laurentiu Vasiliu, S. Haleh S. Dizaji , Aaron Eberhart, Dumitru Roman and Radu Prodan

,

Journal article accepted: Cost modelling and Optimisation for Cloud: A Graph-based Approach

Authors: Akif Quddus Khan, Mihhail Matskin, Radu Prodan, Christoph Bussler, Dumitru Roman, Ahmet Soylu

Journal of Cloud Computing: https://journalofcloudcomputing.springeropen.com/

Abstract: Cloud computing has become popular among individuals and enterprises due to its convenience, scalability, and flexibility. However, a major concern for many cloud service users is the rising cost of cloud resources. Since cloud computing uses a pay-per-use model, costs can add up quickly, and unexpected expenses can arise from a lack of visibility and control. The cost structure gets even more complicated when working with multi-cloud or hybrid environments. Businesses may spend much of their IT budget on cloud computing, and any savings can improve their competitiveness and financial stability. Hence, an efficient cloud cost management is crucial. To overcome this difficulty, new approaches and tools are being developed to provide greater oversight and command over cloud computing expenses. In this respect, this article, presents a graph-based approach for modelling cost elements and cloud resources and a potential way to solve the resulting constraint problem of cost optimisation. In this context, we primarily consider utilisation, cost, performance, and availability. The proposed approach is evaluated on three different user scenarios, and results indicate that it could be effective in cost modelling, cost optimisation, and scalability. This approach will eventually help organisations make informed decisions about cloud resource placement and manage the costs of software applications and data workflows deployed in single, hybrid, or multi-cloud environments.

Paper accepted: SEAL-CC: Scalable Latency Evaluation Methodology for Internet-of-Things Services

Authors: Kurt Horvath, Dragi Kimovski, Radu Prodan, Bernd Spiess, Oliver Hohlfeld

Venue: 14th International Conference on Internet of Things (IoT 2024); Oulu, Finland, 19-22 November, https://iot-conference.org/iot2024

Abstract: Traditional network measurement campaigns suffer from the lack of control over network infrastructure and the inability to evaluate communication performance directly, especially for the placement of highly distributed Internet of Things (IoT) services. In response, we propose a novel Scalable Latency Evaluation Methodology for the Computing Continuum (SEAL-CC). SEAL-CC extends beyond short-term evaluations by capturing the long-term responsiveness of networks supporting IoT services on the computing continuum. It organizes and evaluates a network of nodes, offering insights for optimized IoT service placement in urban and international settings. Our contributions include a novel evaluation methodology tailored for IoT services over the computing continuum, a comprehensive framework for transparent network evaluation using distributed Internet measurement platforms, and a real-life case-study validation with recommendations for IoT service placement.

Paper accepted: Graph Sampling Quality Prediction for Algorithm Recommendation

Authors: Haleh Dizaji, Reza Farahani, Dragi Kimovski, Joze Rozanec, Ahmet Soylu, Radu Prodan

Venue: 31st IEEE International Conference on High Performance Computing, Data, and Analytics; Bengaluru, India,  18-21 December

https://www.hipc.org

Abstract: The increasing size of graph structures in real-world applications, such as distributed computing networks, social media, or bioinformatics, requires appropriate sampling algorithms that simplify them while preserving key properties. Unfortunately, predicting the outcome of graph sampling algorithms is challenging due to their irregular complexity and randomized properties. Therefore, it is essential to identify appropriate graph features and apply suitable models capable of estimating their sampling outcomes. In this paper, we compare three machine learning (ML) models for predicting the divergence of five metrics produced by twelve node, edge, and traversal-based graph sampling algorithms: degree distribution (D3), clustering coefficient distribution (C2D2), hop-plots distribution (HPD2) (including the largest connected component (HPD2C)), and execution time. We use these prediction models to recommend suitable sampling algorithms for each metric and conduct mutual information analysis to extract relevant graph features. Experiments on six large real-world graphs from three categories (scale-free, power-law, binomial) demonstrate an accuracy under 20% in C2D2 and HPD2 prediction for most algorithms despite the relatively high similarity displacement. Sampling algorithm recommendations on ten real-world graphs show higher hits@3 for D3 and

C2D2 and comparable results for HPD2 and HPD2C compared to the K-best baseline method accessing true empirical data. Finally, ML models show superior runtime recommendations compared to baseline methods, with

hits@3 over 86% for synthetic and real graphs and hits@1 over 60% for small graphs. These findings are promising for algorithm recommendation systems, particularly when balancing quality and runtime preferences.

 

,

Reza Farahani gave a tutorial at the 12th European Workshop on Visual Information Processing 2024 in Geneva, Switzerland

Dr. Reza Farahani (University of Klagenfurt, Austria) and Dr. Vignesh V Menon (Fraunhofer HHI, Germany) presented a joint tutorial titled ‘Latency- and Energy-Aware Video Coding and Delivery Streaming Systems’ at the 12th European Workshop on Visual Information Processing (EUVIP 2024) on September 8.

Abstract: This tutorial introduces modern performance and energy-aware video coding and content delivery solutions and tools, focusing on popular video streaming applications, i.e., VoD and live streaming. In this regard, after introducing fundamentals of modern video encoding and networking paradigms, we introduce modern solutions systems, using per-title encoding, per-scene encoding, virtualized and software networks, edge computing, overlay networks such as Content Delivery Networks (CDNs) and/or Peer-to-Peer (P2P) paradigms to provide latency and energy-efficient VoD and live HAS streaming. Furthermore, the tutorial also presents our tools, software, datasets, and testbeds to demonstrate our latest achievements and share practical insights for researchers, engineers, and students who want to improve conversational streaming or even test such techniques for immersive video sequences (e.g., tile-based 360-degree VR) with a focus on latency, economic cost, and energy.

,

Paper accepted for 20th International Conference on Network and Service Management (CNSM 2024)

Title: High Complexity and Bad Quality? Efficiency Assessment for Video QoE Prediction Approaches

Authors: Frank Loh, Gülnaziye Bingöl, Reza Farahani, Andrea Pimpinella, Radu Prodan, Luigi Atzori, Tobias Hoßfeld

Venue: 20th International Conference on Network and Service Management (CNSM 2024)

Abstract:  In recent years, video streaming has dominated Internet data traffic, prompting network providers to ensure high-quality streaming experiences to prevent customer churn. However, due to the encryption of streaming traffic, extensive network monitoring by providers is required to predict the streaming quality and improve their services. Several such prediction approaches have been studied in recent years, with a primary focus on the ability to determine key video quality degradation factors, often without considering the required resources or
energy consumption. To address this gap, we consider existing methods to predict key Quality of Experience (QoE) degradation factors from the literature and quantify the data that have to be monitored and processed for video streaming applications. Based on this, we assess the efficiency of different QoE degradation factor prediction approaches and quantify the ratio between efficiency and the achieved prediction quality. In this context, we identify significant disparities in the efficiency, influenced by data requirements and the specific prediction approach, and finally by the resulting quality. Consequently, we provide insights for network providers to choose the most appropriate method tailored to their specific requirements.

, ,

CERCIRAS Summer School @ AAU – 25.-30.08.2024

CERCIRAS (CA19135), short for Connecting Education and Research Communities for an Innovative Resource Aware Society is a 4-year long COST Action, started at the end of September 2020 and now nearing completion. One of the highlights of CERCIRAS has been its yearly Training School, a week-long residential school designed for current and future PhD students with research interests falling within the thematic scope of the Action and open to other profiles, including industry professionals. While perturbed by unfortunate overlap with the outbreak of the COVID pandemic, CERCIRAS managed to execute three consecutive Training Schools successfully: mid-September 2022 in Split (HR); early September 2023 in Riga (LV); and late August 2024 in Klagenfurt (AT). All editions of the CERCIRAS Training School follow a common structure: 4 selected lecture topics, which alternate frontal lessons and assisted hands-on works; 25 to 35 participants from as many participating countries as possible, with rich diversity for provenance, seniority level, and research focus; a rich portfolio of social activities.

The Training School places a considerable burden on the local hosts, which includes securing comfortable and affordable accommodation for a large troop, providing a modern and spacious lecture hall for all lectures and labs, with refreshments for the two daily breaks and nearby canteens for lunch, and arranging exciting options for a whole-afternoon diversion.

,

6G Continuum Kick-Off Meeting and Research Collaboration

Dragi Kimovski recently visited Mother Theresa University in Skopje as part of the OeAD 6G Continuum project, which focuses on developing middleware for Artificial Intelligence over 6G networks. This visit is the next step in the collaboration between the involved institutions. During the stay, the kick-off meeting for the 6G Continuum project took place, bringing together researchers from both institutions to discuss the project’s goals and clarify the the research activities in terms of supporting AI traning and inference in the Edge over 6G networks. In addition to the research focus, this visit also helped to strengthen the ongoing partnership between the institutions. They explored new opportunities for collaboration in teaching and research, including student and faculty exchanges through Erasmus and CEEPUS programs.