Science AI Journal
Computer Science · Research topic

Open research questions in Cloud Computing and Resource Management

36 unresolved questions extracted from the limitations and future-work sections of 414 Cloud Computing and Resource Management papers in our library. Each links back to the study that raised it.

What the literature leaves open

  • HPC data ingestion is hampered by bandwidth bottlenecks and the limits of traditional methods like FCFS. This paper proposed and evaluated two value-driven auction mecha- nisms, Greedy Value Density and VCG Knapsack, which dynamically allocate ingestion bandwidth. Simulations con- firmed auctions significantly outperform FCFS, reducing transfer completion delays by over 80% in contended con- ditions and capturing over 100% more total scientific value under high load, while also enhancing service predictabil- ity and transfer success rates. A key trade-off emerged: the Greedy auction offers lower computational overhead, while VCG provides stronger truthfulness incentives at a higher computational cost. Value-driven auctions offer a promising paradigm for faster data turnaround, improved system effi- ciency, and fairer resource access in HPC. Future work will focus on validating these mechanisms across other diverse HPC facilities to ensure generalizability, and developing hy- brid approaches—such as utilizing the Greedy algorithm for rapid baseline allocation and dynamically triggering VCG during peak contention—prior to operational deployment.

    XOR Bidding and Knapsack Formulations for HPC Network Resource Allocation · 2026 · DOI

  • Task scheduling in cloud computing based on meta- heuristics: Review, taxonomy, open challenges, and future trends. Additionally, this study was limited to the BestGap scheduler and two datasets. This will provide a more reliable estimate of model performance over time and enable fair comparisons between architectures, even with extremely limited data. The choice of predictive architecture must be driven by the structure of the target: sharp zeros require two-stage decomposition, while continuous targets may suffice with simpler models—especially when data is scarce.

    From Wait Time to Slowdown: A Hurdle Framework for Performance-Centric Metric Prediction in HPC Clusters · 2026 · DOI

  • Fixed parameters Single-objective focus 2026 GA for fog-cloud deployment maximization Latency/energy/bandwidth minimization 2026 EPPO game theory framework Cost-effective scientific scheduling Early convergence 2026 HLFO + Q-Learning 2026 MCT-initialized…

    Hybrid Genetic Algorithm and Water Wave Optimization Approach for QoS Aware Multi Objective Task Scheduling and Load Balancing in Cloud Environments · 2026 · DOI

  • To this end, future research should focus on comprehensively analysing the performance characteristics of serverless functions implemented in different programming languages across various serverless platforms. Additionally, the support for specialised hardware such as Graphics Process- ing Unit (GPU) and Field-Programmable Gate Array (FPGA) remains an open question that could further have a significant impact on resource configuration decisions for specialised workloads. Therefore, function configuration-aware scheduling algorithms that are also adaptable to predict and mitigate potential co-location issues should be explored in future studies. Serverless Computing: Current Trends and Open Problems.

    Dynamic Function Configuration and its Management in Serverless Computing: A Taxonomy and Future Directions · 2026 · DOI

  • This research introduced the hybrid scheduling framework MO-NSGAQ to address the challenges of task scheduling in federated cloud settings. It combines NSGA-II with Q-learning. By combining multi-objective optimization with RL, MO-NSGAQ effectively balances trying new things and learning, unlike traditional schedulers that rely on fixed rules or only evolutionary techniques. The proposed framework dynamically adapts to changing workloads, resource restrictions, and inter-cloud heterogeneity ARTICLE IN PRESS ARTICLE IN PRESS ACCEPTED MANUSCRIPT owing to this hybrid combination. The proposed framework provided better results when compared with existing approaches, including MOPSO, MOABCQ, Max-Min, and FCFS. Comprehensive simulation runs are conducted on three different datasets, such as synthetic workloads, Google Cloud traces, and real-time IoT workloads. With high throughput, MO-NSGAQ improved significantly in other metrics such as makespan reduction, cost efficiency, load balancing, and resource utilization. Adaptability and intelligent feedback are important in complicated, multi- provider environments, where the performance increases are most considerable. A wide variety of distributed computing systems benefit from MO-NSGAQ's design, not just federated cloud scheduling. It can be used in edge-cloud hybrid systems, smart grid task management, fog computing, and managing sustainable data centres, where there are often conflicting goals and situations. It is a flexible option for intelligent infrastructure of the future because of its learning-driven flexibility and Pareto-optimal reasoning. In the future there are several ways this study may have a greater impact, including deep reinforcement learning integration, SLA-aware scheduling, energy and carbon-aware optimisation, real-time deployment in federated testbeds, and multi-agent Q-learning extensions. Future work will also include comparisons with existing DRL-based schedulers, as well as obvious evaluation of inter-cloud communication overhead, SLA violation rates, and scheduling decision latency to assess real-world deployability.

    Federated multi-cloud task scheduling with load balancing using multi-objective NSGA-II and reinforcement learning · 2026 · DOI

  • While our serverless communicator design demonstrates performance comparable to traditional EC2 infrastructure, several limitations remain. First, Cylon currently implements only about 30% of the operators supported by Pandas, limiting its applicability for certain data processing workflows. Second, AWS Lambda imposes an upper limit of fifteen minutes for function invocations, which, combined with our BSP-styled architecture, restricts the applicability of data parallel processing for large datasets that require extended processing time. Third, the lack of checkpointing and fault tolerance mechanisms limits the ability to recover from failures or time-constrained execution boundaries in serverless environments. To address these limitations, we would like to converge Cylon and Pandas operators using a comparative, high-performance API built using Cython. For example, we have begun to design the windowing operator and plan to complete that work soon. Utilizing the cost model (detailed by Perera in “Towards Scalable High Performance Data Engineering Systems”), we can evaluate data parallel operator patterns to identify bottlenecks in communication operations. We believe we can significantly improve performance by designing other algorithms with lower latency (Perera et al., 2023). Additionally, related to communicator support, we would also like to add libfabric as a communicator based on implementations across cloud services based on GPU hardware supported by cloud providers. To address the time limitation and lack of fault tolerance, we want to design a checkpointing feature similar to that implemented in the Twister2 library. This will allow for the recovery of unfinished executions based on upper-limit time constraints imposed by AWS Lambda. This will also provide general fault tolerance for MPI workloads outside of the serverless use case. Lastly, we would like to investigate real-world applications and detail the performance and implications of serverless as a possibility for deep learning and machine learning inference on clouds. For example, we would like to implement hydrology and earthquake prediction deep learning on both serverful and serverless AWS to demonstrate the applicability and potential scale offered by cloud providers like as AWS.

    Combining serverless and high-performance computing paradigms to support ML data-intensive applications · 2026 · DOI

  • Future work may focus on incorporating richer contextual features, exploring uncertainty-aware prediction, and extending evaluation to multiple datasets in order to improve generalizability and better capture extreme-tail be- havior. Error analysis also reveals that the proposed method still underestimates the largest spikes, indicating that accurate prediction of rare extreme jobs remains an open chal- lenge. 500 that the regime-aware model consistently achieves the best mean RMSE and mean R² across temporal folds, suggesting that its advantage is not limited to a single data partition.

    Log-Transformed Regime-Based Prediction of Cloud Job Length Using Machine Learning · 2026 · DOI

  • This review synthesized findings from 21 studies and also noted several limitations. spanning multiple databases, subscription-only literature. longer time windows, and 7.3 Future research directions Future work should extend the existing frameworks to heterogeneous and dynamic environments by incorporating adaptive load balancing and dynamic scheduling techniques.

    Scientific and practical basis for implementing machine learning on a parallel computing cluster: a systematic mapping review · 2026 · DOI

  • The CNN-Bi-LSTM-Attention architecture uses fixed hyperparameters (CNN kernel sizes, Bi-LSTM hidden dimensions, attention heads) across all experiments. The sensitivity of the model to hyperparameter variations and whether architecture components should be automatically tuned based on workload characteristics or time-series properties is not addressed.

    An Adaptive Deep Learning Framework with Enhanced Attention for Precise Load Forecasting in Cloud Computing Environments · 2026 · DOI

  • The paper demonstrates a 2.5% reduction in total provisioned capacity and 13.9% reduction in overprovisioning, but does not investigate how the safety margin (σ) should be dynamically adjusted based on SLA requirements, workload volatility, or different service tiers in heterogeneous cloud environments.

    An Adaptive Deep Learning Framework with Enhanced Attention for Precise Load Forecasting in Cloud Computing Environments · 2026 · DOI

  • Cross-dataset generalization was tested on Google Cluster Trace and synthetic anomaly injection (recall >0.90), but evaluation on domain-specific cloud workload datasets (e.g., Azure VM traces, Amazon EC2 workloads, Alibaba cluster data) with different temporal characteristics, seasonality patterns, and resource utilization profiles has not been performed.

    An Adaptive Deep Learning Framework with Enhanced Attention for Precise Load Forecasting in Cloud Computing Environments · 2026 · DOI

  • The model achieves 94.2% anomaly detection recall (162/172) on the IoT gateway test set but the paper does not specify how the labeling and ground truth for anomalies were established, whether domain experts were involved in annotation, or how label consistency was validated across the 172 anomaly instances.

    An Adaptive Deep Learning Framework with Enhanced Attention for Precise Load Forecasting in Cloud Computing Environments · 2026 · DOI

  • The attention mechanism's contribution to load forecasting was demonstrated through ablation study (34.3% MSE reduction), but the paper does not analyze which specific time steps or temporal patterns the attention weights prioritize during peak loads, load spikes, or anomalous workload shifts in cloud computing environments.

    An Adaptive Deep Learning Framework with Enhanced Attention for Precise Load Forecasting in Cloud Computing Environments · 2026 · DOI

  • All computational cost evaluations were conducted on a single Dell XPS 15 laptop (CPU-only, <5% GPU utilization) with 16GB RAM. The framework's scalability and practical deployment efficiency on edge devices with constrained memory (<4GB), on GPU-accelerated cloud nodes, and in distributed inference scenarios with multiple forecasting agents has not been investigated.

    An Adaptive Deep Learning Framework with Enhanced Attention for Precise Load Forecasting in Cloud Computing Environments · 2026 · DOI

  • The proposed CNN-Bi-LSTM-Attention model was evaluated on a self-collected IoT gateway dataset of 8,622 samples and validated only on Google Cluster Trace and synthetic anomaly injection (Gaussian noise + spikes). The model's performance on real-world cloud workloads from heterogeneous data centers with diverse hardware configurations, workload patterns (batch processing, streaming, interactive), and multi-tenant resource contention scenarios remains unexplored.

    An Adaptive Deep Learning Framework with Enhanced Attention for Precise Load Forecasting in Cloud Computing Environments · 2026 · DOI

  • The paper claims the QPU identifies the global optimum four times faster than classical search strategies but does not provide detailed comparative analysis or theoretical justification for this claim.

    A Quantum-Inspired Evolutionary Framework for Multi-Objective Task Scheduling in Sustainable Cloud Computing Environments · 2026 · DOI

  • The rapid growth of CDCs worldwide has led to enormous amount of electric energy consumption, thereby increasing CO2 emissions. This work focuses on the cloud IaaS platform in two ways. First, the overall electric energy consumption, which directly impacts the running cost of the CDC, is minimized. Thus, this study provides a real environment of cloud computing for growing its industries all over the world. Second, SLA violation, number of reactivated hosts, and number of VM migrations, are (a) (b) Fig.

    An adaptive heuristic for managing energy consumption and overloaded hosts in a cloud data center · 2018 · DOI

  • However, the suitability of emerging AI workloads for serverless remains insufficiently explored.

    Agentic Workflows are Serverless Applications, so deploy them that way! · 2026 · DOI

  • (2) Agentic AI workloads have not yet been characterized under the serverless context.

    Agentic Workflows are Serverless Applications, so deploy them that way! · 2026 · DOI

  • , "Survey on explainable AI: Techniques, challenges and open issues," Expert Systems with Applications, Available: vol.

    Enabling Real-Time Transaction Processing in Distributed Cloud Systems: Architectural, Operational, and Societal Dimensions · 2026 · DOI

  • Hossain, “A comprehensive examination of load balancing a environments: algorithms systematic review, comparative analysis, taxonomy, open challenges, and future trends,” Iran Journal of Computer Science, vol.

    Optimizing Cloud Resource Scheduling: A Comparative Study of ACO, PSO, GA, and Hybrid Approaches in CloudSim · 2026 · DOI

  • A limitation of this study is the reliance on TDP-based energy estimation for nodes where direct power measurement was unavailable.

    Energy-Efficient Kubernetes Scheduling for Green Cloud Computing: A Carbon-Aware Workload Placement Framework · 2026 · DOI

  • Identified three major trends in query optimization: tighter feedback loops between optimization and execution, workload-level optimization, and composable architectures for cross-engine collaboration.

    Review of Query Optimization Techniques in Database Management Systems · 2026 · DOI

  • The practical implementation details of the QUBO transformation for the scheduling problem are not thoroughly discussed or validated.

    A Quantum-Inspired Evolutionary Framework for Multi-Objective Task Scheduling in Sustainable Cloud Computing Environments · 2026 · DOI

  • The paper mentions incorporating Quantum Cryptography elements to protect data transmission between nodes but does not provide detailed methodology or implementation.

    A Quantum-Inspired Evolutionary Framework for Multi-Objective Task Scheduling in Sustainable Cloud Computing Environments · 2026 · DOI

Most-cited papers in Cloud Computing and Resource Management

Most recent work

Find a gap in your own Cloud Computing and Resource Management sub-topic

This page shows what the Cloud Computing and Resource Management literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.

Open the Research Gap Finder →

Related topics in Computer Science

Online Learning and Analytics
363 open questions
AI in Service Interactions
278 open questions
Teaching and Learning Programming
239 open questions
Digital literacy in education
230 open questions
Mobile Learning in Education
183 open questions
Virtual Reality Applications and Impacts
178 open questions
Computational Drug Discovery Methods
126 open questions
Blockchain Technology Applications and Security
121 open questions
36 open questions have been extracted from the limitations and future-work passages of 414 Cloud Computing and Resource Management papers in our library. Each one below links back to the study that raised it, so you can read the original claim in context.

Tools for your next paper

Gaps & Literature SearchFind open questions in any field — 120,000+ open gaps mined from our own library.Pre-CheckIs your paper ready? A Tier 1–5 acceptance probability in seconds.Journal FinderA ranked shortlist of target journals from a 17,500-venue index.Duplicate Publication CheckerCheck prior-publication & salami-slicing risk across 8 sources.Citation GeneratorFreeA DOI or title → APA, MLA, Vancouver, Chicago, BibTeX, RIS. No signup.Graphical Abstract MakerFreeTurn your findings into a colorblind-safe graphical abstract.

Compare the category — Honest roundups of the AI research tools, ours listed alongside the alternatives.

Command palette

Jump anywhere, run any action.