Pith. sign in

REVIEW 4 major objections 4 minor 63 references

Sentinel: Scheduling Live Streams with Proactive Anomaly Detection in Crowdsourced Cloud-Edge Platforms

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Scheduling live streams with proactive anomaly detection cuts ineffective scheduling by 70% and raises revenue by 74%.

desk verdict The P2S scheduling architecture is a genuine new idea, but the 74% revenue gain is an unvalidated proxy, so the quantitative claims need a lot more support. read the letter →

arxiv 2505.23347 v1 pith:YDWVPLAY submitted 2025-05-29 cs.LG

Yuting Li , Shaoyuan Huang , Tengwen Zhang , Cheng Zhang , Xiaofei Wang , Victor C.M. Leung This is my paper · ORCID
classification cs.LG
keywords livestreamingcloud-edgecomputingschedulinganomalydetectionrevenueoptimizationpre-poststrategypoolvariationalGRU
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that existing schedulers for crowdsourced cloud-edge platforms underperform because they ignore anomalies, which are common and costly. It proposes Sentinel, a two-stage Pre-Post-Scheduling framework that moves anomaly detection and revenue-optimal strategy generation into a pre-scheduling stage, then uses a strategy pool to match arriving requests quickly. On five days of real-world CCP data, Sentinel reports a 70% reduction in anomaly-triggered ineffective scheduling, a 74% average revenue improvement over the platform's original heuristic, and a 2.0x scheduling speedup. If correct, this makes anomaly-aware scheduling practical at the scale and time-sensitivity of live streaming.

What carries the argument

The key machinery is the Pre-Post-Scheduling (P2S) paradigm, which separates expensive preparation from real-time decisions. In the pre-scheduling stage, a two-step device anomaly detector (rule-based variance/correlation checks followed by a variational GRU with Gaussian-mixture latent variables and Gumbel-softmax sampling) and a service-effect predictor (TCN-GCN for request volume, cross-attention decoder for revenue and anomaly status) generate a revenue-maximizing strategy by solving a relaxed linear program with branch-and-cut. The strategy is stored in a strategy pool; the post-scheduling stage matches arriving requests to the stored strategy and falls back to the original heuristic only for prediction errors. The Gaussian revenue-efficiency function $F_r(U) = e^{-(U-U_{opt})^2/(2\sigma^2)}$, with $U_{opt}$ from the 80th percentiles of latency and error rate, is the objective that ties utilization to revenue, making the optimization tractable and the decoupling of stages possible.

What would settle it

Run Sentinel on a platform where actual SLA fines and paid revenue are logged, and compare the realized revenue improvement against the reported 74%: if the Gaussian-derived schedule increases utilization without increasing realized revenue, the revenue model is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a proactive, anomaly-aware scheduling architecture can both avoid service disruptions and increase revenue in a real crowdsourced cloud-edge platform. The authors analyze real traces and find that device and service anomalies, especially device failures, cause most SLA fines. Sentinel detects device anomalies with a two-step detector (fast rules plus a variational GRU with a Gaussian-mixture prior), predicts service anomalies and request revenue with a TCN-GCN plus cross-attention model, and solves a revenue-maximization problem whose objective is a Gaussian revenue-efficiency function of utilization. The resulting strategies fill a strategy pool used at runtime, with the platform's own heuristic as a fallback for unmatched requests. In five-day tests, this reduces the frequency of both anomaly types to below 0.05 and improves average revenue by 74% over the original heuristic.

Load-bearing premise

The Gaussian revenue-efficiency function $F_r(U) = e^{-(U-U_{opt})^2/(2\sigma^2)}$, with $U_{opt}$ fixed at the 80th percentile of latency and error rate, is an assumed model of how utilization translates into real revenue after SLA fines; if that curve does not match reality, the reported 74% revenue improvement is an improvement in a proxy, not in actual money.

Editorial extensions

If this is right

  • If the reported results hold, anomaly detection can be moved out of the real-time scheduling path, so scheduling latency becomes nearly independent of detector complexity.
  • The strategy-pool approach implies that each incoming request only needs a look-up, so the scheduling decision time is small and bounded.
  • The Gaussian revenue-efficiency function gives operators a single utilization target around which to balance load, which can be updated as new latency and error data arrive.
  • Because the pre-scheduled strategy is rounded from a relaxed linear program, the framework offers a computationally scalable way to approximate a hard integer optimization.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension is to replace the hand-specified Gaussian revenue function with a learned or measured revenue function from real SLA fine data; the paper does not report how sensitive the 74% revenue gain is to the exact shape of $F_r(U)$ or to the choice of $\sigma$ and $U_{opt}$.
  • The P2S separation could generalize beyond live streaming to other edge-cloud scheduling tasks where anomaly-prone devices and time-sensitive decisions coexist, such as real-time machine learning inference or IoT data processing.
  • The paper's anomaly rate of 16.1% and its threshold tuning suggest the results may depend on the reliability characteristics of this particular CCP; applying Sentinel to a much more stable or much more unstable platform would likely require re-tuning the detector thresholds.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes Sentinel, a two-stage pre/post-scheduling framework for live streaming workloads on crowdsourced cloud-edge platforms (CCPs). In the pre-scheduling stage, Sentinel detects device anomalies with a rule-based detector followed by a mixture-VGRU deep model, predicts request volumes and per-request service effects with a TCN-GCN plus cross-attention model, and solves a revenue-maximizing assignment problem to build a strategy pool. At request time, the post-scheduling stage matches actual requests to the pre-generated strategies and falls back to a heuristic for unmatched requests. Using real traces from a Chinese CCP, the authors report that Sentinel reduces anomaly frequency by 70%, improves revenue by 74%, and doubles scheduling speed relative to several baselines. The core scheduling optimization is a relaxed approximation of an NP-hard integer nonlinear program, and the revenue objective relies on a Gaussian 'revenue efficiency' function.

Significance. If the reported results hold, the paper makes a useful practical contribution: it is, to my knowledge, the first to explicitly integrate proactive anomaly detection into CCP scheduling for live streaming, and the two-stage paradigm is a sensible way to move expensive detection/optimization off the critical path. The measurement study of real CCP anomalies and the mixture-VGRU device detector are reasonable components, and the evaluation is grounded in substantial real-world datasets. However, the headline revenue claim rests on a single unvalidated Gaussian revenue surrogate that is used both as the optimizer's objective and as the evaluation metric, and the Origin baseline is not run under the same conditions as Sentinel. These issues are load-bearing for the paper's central quantitative claims, so the contribution cannot be fully assessed without additional validation.

major comments (4)
  1. [II-C, Eq. (1), Eq. (10), Fig. 8(c)] The revenue efficiency function F_r(U) = exp(-(U-U_opt)^2/(2σ^2)) is used as the optimization objective in Eq. (10), and the revenue reported in Fig. 8(c) appears to be computed from the same model. The parameter σ is never specified, and no validation of Eq. (1) against actual CCP revenue net of SLA fines is provided. Consequently, the claimed 70%, 134%, 17%, and 76% revenue improvements may be improvements in a proxy rather than in real monetary revenue. I request that the authors either (a) fit and validate Eq. (1) against ground-truth revenue/SLA-fine data withheld from the fitting, including a reported σ and goodness-of-fit, or (b) report revenue using an independently measured metric and show that the ranking of methods is robust to σ and to alternative revenue models.
  2. [IV-A2, Baselines, Fig. 8] The Origin baseline is not evaluated in the same simulation as Sentinel: the text states that for Origin the authors 'do not perform additional replication but only record the relevant metrics under the real CCP.' Because Sentinel is run on a testbed built from CCP clusters, the comparison against Origin mixes environmental differences with algorithmic differences. Please rerun Origin in the same simulated testbed, or clearly label it as a historical reference and avoid drawing quantitative improvement claims from that comparison.
  3. [III-A3, Eqs. (10)-(14)] The road from the NP-hard integer nonlinear program to the solved linear program involves four approximations: dropping the location dimension (x_e_m_i -> x_e_i), replacing the binary service anomaly S with a sampled expectation, applying a secant outer approximation, and relaxing/rounding the integer variables. No bound or empirical optimality gap is reported for any of these steps. Since the final strategy is the rounded solution, it is unclear how much of the reported revenue gain is attributable to anomaly detection versus how much is lost or gained through the approximations. Please provide a small-scale comparison against exact enumeration or an upper bound, and report the sensitivity of the final revenue to each approximation stage.
  4. [IV-B, Fig. 8] The evaluation spans only five days and no variance or confidence intervals are reported. The anomaly-frequency and revenue differences in Fig. 8(a) and Fig. 8(c) could be within sample noise, especially for a system with roughly 16% observed anomaly frequency. Please include variability across runs or across time periods (e.g., bootstrap over the five days or multiple simulated traces) and clarify what the y-axis 'Revenue' in Fig. 8(c) measures in units.
minor comments (4)
  1. [III-A1, Eq. (4)] The sentence following Eq. (4) says an observation is classified as anomalous if the anomaly score is below η, but Eq. (4) states the opposite (score > η). Please correct the text to match the equation.
  2. [Throughout, Eqs. (2)-(10)] The symbol N is used for the number of workload dimensions in Eq. (2) and also for the number of request categories in R_{t,m}; later I is used for categories in Eq. (10). This notation conflict makes the formulation harder to follow. Please harmonize the notation.
  3. [IV-B] The claim that Sentinel 'reduces ineffective scheduling by an average of 70%' would benefit from an explicit definition of ineffective scheduling; as written, it is not clear whether this is the same as the anomaly-frequency reduction in Fig. 8(a).
  4. [I, Abstract] The phrase '2.0×the scheduling speed' is missing a space; it should read '2.0× the scheduling speed'.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline 74% revenue improvement is evaluated with the same Gaussian revenue surrogate that Sentinel optimizes, so the revenue claim is partly circular.

  1. self definitional [Sec. II-C Eq. (1); Sec. III-A3 Eq. (10); Sec. IV-B3 Fig. 8(c)]
    "we introduce a Gaussian function to model F_r(·) and implement smoothing as: R(U) = e^{-(U-U_opt)^2/(2σ^2)} (1) ... The pre-scheduling problem can be defined as: max Σ_{e=1}^E F_r(Σ_{m,i} x_{m,i}^e * A_{e,m,i} * S_{e,m,i} / B_e) * D_e. (10) ... Sentinel achieves an enhancement of 70%, 134%, 17%, and 76% in the average revenue comparing to original heuristic method, GP, Greedy and MF, respectively."

    Eq. (10) maximizes exactly the F_r defined in Eq. (1), and the paper's only revenue metric is Eq. (1); Fig. 8(c) is reported as 'revenue' without any independent validation against SLA fines or actual money. Thus the Fig. 8(c) comparison evaluates the same hand-defined Gaussian function that Sentinel optimizes. Any scheduler that keeps utilization near U_opt receives F_r ≈ 1 by construction, so the reported 70%/134%/17%/76% gains are differences within the paper's own surrogate, not an independently measured outcome. The anomaly-frequency reduction in Fig. 8(a) is ground-truth based and non-circular, but the headline revenue improvement is partially forced by using the objective function as the evaluation metric.

full rationale

The paper's strongest revenue claim is circular in the specific, mild sense that the optimizer and the evaluator share the same unvalidated revenue-efficiency function. Eq. (1) defines revenue as a Gaussian function of utilization; Eq. (10) maximizes that exact function; and Sec. IV-B3 reports 'revenue' improvements without showing that the y-axis of Fig. 8(c) is anything other than Eq. (1). Since U_opt is derived from data and σ is left unspecified, the surrogate is not independently confirmed, so the 74% figure is an improvement in the paper's own proxy rather than demonstrated monetary revenue. The anomaly-frequency reduction (70%) is based on real anomaly counts and is not circular. The device-anomaly detector, service-effect predictor, and scheduling formulation are internally consistent and have independent empirical components; no load-bearing self-citation or imported uniqueness theorem was found. The paper would be non-circular if revenue were validated against actual SLA/fine accounting data; as written, the revenue claim reduces partly by construction.

Assumptions & free parameters 7 free parameters · 4 assumptions · 1 invented entities

The central claim rests on several hand-set or data-fitted parameters (sigma, U_opt, detection thresholds, cluster counts) and on an ad hoc Gaussian revenue model that is used both for optimization and evaluation. No new physical entities are introduced; the only invented conceptual object is the revenue metric itself.

free parameters (7)
  • sigma in revenue efficiency F_r(U)
    Controls the width of the Gaussian revenue function (Eq. 1). No value or fitting procedure is given in the paper, but it strongly affects the objective and therefore the schedule.
  • U_opt (optimal utilization) = min(ULat80, UErr80)
    Computed from 80th percentiles of historical startup latency and error rates (Eq. 12). It is a data-derived threshold used in both the revenue model and the utilization constraint (Eq. 11c).
  • theta_v and theta_r (rule-based detection thresholds) = 0.4
    Set to 0.4; the paper's robustness study (Fig. 11) picks theta=0.4 and eta=0.3 as the best point for F1 score.
  • eta (learning-based anomaly threshold) = 0.3
    Threshold on reconstruction error in Eq. (4); chosen via the F1 sweep in Fig. 11.
  • k (number of request clusters) = 29
    Chosen for KMeans clustering of request categories in Sec. IV-A1; no rationale is given for this specific value.
  • T (sliding window length) = 12
    Window length for the multivariate workload series in rule-based detection; set in Sec. IV-A1.
  • K (number of mixture components in VGRU)
    Not specified in the paper; appears in Eq. (5) as the number of Gaussian components.
assumptions (4)
  • standard math Standard linear algebra, probability, and deep learning results underlying GCN, TCN, VGRU, and Gumbel-softmax.
    Used without proof throughout Sec. III-A.
  • domain assumption SLA violations can be categorized into service, artificial, device, and orchestration failures, and their rates are stable enough to learn from five days of logs.
    Sec. II-A introduces this taxonomy, which drives the anomaly detection design.
  • domain assumption Service anomaly probability depends primarily on peak versus off-peak timing and inter-region versus intra-region routing.
    Supported by Fig. 3(c), but treated as a fixed feature set in the cross-attention model.
  • ad hoc to paper The Gaussian function (Eq. 1) with a single U_opt captures revenue-versus-utilization trade-offs across all heterogeneous servers.
    This is a modeling choice not derived from a first-principles model of revenue.
invented entities (1)
  • Gaussian revenue efficiency metric F_r(U)
    purpose: Serves as both the scheduling objective (Eq. 10) and the revenue evaluation metric (Sec. IV-B).
    It is introduced in Sec. II-C to smooth observed nonlinear throughput-revenue behavior, but is not validated against monetary revenue or an independent dataset.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Sentinel: Scheduling Live Streams with Proactive Anomaly Detection in Crowdsourced Cloud-Edge Platforms." pith.science (2026). https://pith.science/paper/YDWVPLAY

@misc{pith2026250523347,
  author       = {Pith},
  title        = {Pith review of: Sentinel: Scheduling Live Streams with Proactive Anomaly Detection in Crowdsourced Cloud-Edge Platforms},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YDWVPLAY}},
  note         = {Machine review of arXiv:2505.23347}
}
read the original abstract

With the rapid growth of live streaming services, Crowdsourced Cloud-edge service Platforms (CCPs) are playing an increasingly important role in meeting the increasing demand. Although stream scheduling plays a critical role in optimizing CCPs' revenue, most optimization strategies struggle to achieve practical results due to various anomalies in unstable CCPs. Additionally, the substantial scale of CCPs magnifies the difficulties of anomaly detection in time-sensitive scheduling. To tackle these challenges, this paper proposes Sentinel, a proactive anomaly detection-based scheduling framework. Sentinel models the scheduling process as a two-stage Pre-Post-Scheduling paradigm: in the pre-scheduling stage, Sentinel conducts anomaly detection and constructs a strategy pool; in the post-scheduling stage, upon request arrival, it triggers an appropriate scheduling based on a pre-generated strategy to implement the scheduling process. Extensive experiments on realistic datasets show that Sentinel significantly reduces anomaly frequency by 70%, improves revenue by 74%, and doubles the scheduling speed.

Figures

Figures reproduced from arXiv: 2505.23347 by the authors.

Figure 1
Figure 1. Crowdsourced Cloud-edge service Platform. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Server utilization analysis. (ii) Scheduling Invalid. While previous scheduling opti￾mization efforts have been considered effective, they have all overlooked anomalous disruptions, which can only get unre￾alistic results because the platform will not remain perfectly healthy all the time as people ideally. Therefore, it is crucial to identify various anomalies within the unstable CCPs, develop specialized technolog… view at source ↗
Figure 3
Figure 3. Analysis of the anomalies. yellow segments, the memory usage and disk usage exhibit normal fluctuations with consistent tendency. Nevertheless, there are sudden changes with inconsistent tendency of each series in the blue segments, which can be attributed to corre￾lation disturbances caused by device anomaly. Consequently, it is vital to consider both non-stationary temporal character￾istics (fluctuations) and corr… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Resource revenue metric. However, it can hardly be used directly as the CCPs revenue metric due to its discontinuities and non-universality (servers with different hardware conditions exhibit varying throughput levels and prices). To address these issues, we design a u…
Figure 5
Figure 5. Figure 5: The system overview of Sentinel. extending the native scheduling into a Pre-Scheduling stage and a Post-Scheduling stage with Strategy Pool. To address the significant time costs of detection in real￾time scheduling, Sentinel conducts anomaly detection during the pre-s…
Figure 6
Figure 6. Figure 6: Two stage anomaly detector. However, traditional VGRU models use a single Gaussian distribution for latent variables zt,n, making it challenging to capture diverse patterns in heterogeneous servers and unsuit￾able for large-scale CCPs. To address this, we incorporate a…
Figure 7
Figure 7. Figure 7: Service effect prediction. them through the equation A˜ = A + IN , D˜ ii = P j A˜ ij , and IN is the identity matrix. Given our use of the standard TCN, for brevity, we adopt TCN to represent the standard computational process inherent in the TCN. Step 2: Service Effec…
Figure 8
Figure 8. Figure 8: Scheduling performance comparison1 . by an average of 70% compared to other baselines. Sentinel directs requests to reliable servers by filtering out those likely to cause scheduling issues based on anomaly detection. 2) Sentinel Yields More Balanced Utilization [PITH…
Figure 9
Figure 9. Figure 9: Average scheduling time consumption. reduces the scheduling time consumption by 1.5×, 1.6×, and 2.8× compared to GP, Greedy, and MF, respectively. Current state-of-the-art scheduling methods (e.g., [22], [23]) fail to fully exploit predictive capabilities, often tailor…
Figure 11
Figure 11. Figure 11: Variation in anomaly detection performance with different settings. [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 62 canonical work pages

  1. [1]

    Live streaming statistics 2023: Facts about live streaming in the u.s

    ZIPPIA, “Live streaming statistics 2023: Facts about live streaming in the u.s.” 2023. [Online]. Available: https://www.zippia.com/advice/live- streaming-statistics/

    work page 2023

  2. [2]

    Ugc vs aigc- an in-depth analysis,

    OPPORTURE, “Ugc vs aigc- an in-depth analysis,” 2023. [Online]. Available: https://www.opporture.org/content-creation/comparing-ugc- and-aigc/

    work page 2023

  3. [3]

    User mapping strategy in multi-cdn streaming: A data-driven approach,

    G. Zhu and W. Gu, “User mapping strategy in multi-cdn streaming: A data-driven approach,”IEEE Internet of Things Journal, pp. 6638–6649, 2021

    work page 2021

  4. [4]

    On Uploading Behavior and Optimizations of a Mobile Live Streaming Service,

    J. Li, Z. Li, Q. Wu, and G. Tyson, “On Uploading Behavior and Optimizations of a Mobile Live Streaming Service,”Proceedings - IEEE INFOCOM, pp. 1299–1308, 2022

    work page 2022

  5. [5]

    Look Ahead at the First-mile in Livecast with Crowdsourced Highlight Prediction,

    C. Zhang, J. Liu, Z. Wang, and L. Sun, “Look Ahead at the First-mile in Livecast with Crowdsourced Highlight Prediction,”Proceedings - IEEE INFOCOM, pp. 1143–1152, 2020

    work page 2020

  6. [6]

    QoE-Aware Resource Allocation for Crowdsourced Live Streaming: A Machine Learning Approach,

    F. Haouari, E. Baccour, A. Erbad, A. Mohamed, and M. Guizani, “QoE-Aware Resource Allocation for Crowdsourced Live Streaming: A Machine Learning Approach,”IEEE International Conference on Communications, vol. 2019-May, 2019

    work page 2019

  7. [7]

    Proactive Video Push for Optimizing Bandwidth Consumption in Hybrid CDN-P2P V oD Systems,

    Y . Zhang, C. Gao, Y . Guo, K. Bian, X. Jin, Z. Yang, L. Song, J. Cheng, H. Tuo, and X. Li, “Proactive Video Push for Optimizing Bandwidth Consumption in Hybrid CDN-P2P V oD Systems,”Proceedings - IEEE INFOCOM, pp. 2555–2563, 2018

    work page 2018

  8. [8]

    Aggcast: Practical cost-effective scheduling for large-scale cloud-edge crowdsourced live streaming,

    R.-X. Zhang, C. Yang, X. Wang, T. Huang, C. Wu, J. Liu, and L. Sun, “Aggcast: Practical cost-effective scheduling for large-scale cloud-edge crowdsourced live streaming,” inProceedings of the 30th ACM International Conference on Multimedia, 2022

    work page 2022

Show all 63 references
  1. [9]

    Cfa: A practical prediction system for video qoe optimization,

    J. Jiang, V . Sekar, H. Milner, D. Shepherd, I. Stoica, and H. Zhang, “Cfa: A practical prediction system for video qoe optimization,” in13th USENIX Symposium on Networked Systems Design and Implementation (NSDI 16), 2016, pp. 137–150

    2016

  2. [10]

    Integration of broadcaster and telco access networks for real time/live events,

    R. Giuliano, F. Mazzenga, and A. Vizzarri, “Integration of broadcaster and telco access networks for real time/live events,”IEEE Transactions on Broadcasting, vol. 66, no. 3, pp. 667–675, 2020

    2020

  3. [11]

    Video caching, analytics, and delivery at the wireless edge: a survey and future directions,

    B. Jedari, G. Premsankar, G. Illahi, M. Di Francesco, A. Mehrabi, and A. Yl ¨a-J¨a¨aski, “Video caching, analytics, and delivery at the wireless edge: a survey and future directions,”IEEE Communications Surveys & Tutorials, vol. 23, no. 1, pp. 431–471, 2020

    2020

  4. [12]

    A contemporary survey on live video streaming from a computation- driven perspective,

    N.-N. Dao, A.-T. Tran, N. H. Tu, T. T. Thanh, V . N. Q. Bao, and S. Cho, “A contemporary survey on live video streaming from a computation- driven perspective,”ACM Comput. Surv., 2022

    2022

  5. [13]

    Unreeling netflix: Understanding and improving multi-cdn movie delivery,

    V . K. Adhikari, Y . Guo, F. Hao, M. Varvello, V . Hilt, M. Steiner, and Z.- L. Zhang, “Unreeling netflix: Understanding and improving multi-cdn movie delivery,” in2012 Proceedings IEEE INFOCOM. IEEE, 2012, pp. 1620–1628

    2012

  6. [14]

    Enhancing the crowdsourced live streaming: A deep reinforcement learning approach,

    R.-X. Zhang, T. Huang, M. Ma, H. Pang, X. Yao, C. Wu, and L. Sun, “Enhancing the crowdsourced live streaming: A deep reinforcement learning approach,” inProceedings of the 29th ACM Workshop on Network and Operating Systems Support for Digital Audio and Video, 2019, pp. 55–60

    2019

  7. [15]

    Resource management for infras- tructure as a service (iaas) in cloud computing: A survey,

    S. S. Manvi and G. Krishna Shyam, “Resource management for infras- tructure as a service (iaas) in cloud computing: A survey,”Journal of Network and Computer Applications, 2014

    2014

  8. [16]

    Containerization and the paas cloud,

    C. Pahl, “Containerization and the paas cloud,”IEEE Cloud Computing, vol. 2, no. 3, pp. 24–31, 2015

    2015

  9. [17]

    Profit-aware resource management for edge computing systems,

    C. Anglano, M. Canonico, and M. Guazzone, “Profit-aware resource management for edge computing systems,” inProceedings of the 1st International Workshop on Edge Systems, Analytics and Networking, 2018, pp. 25–30

    2018

  10. [18]

    Proactive video push for optimizing bandwidth consumption in hybrid cdn-p2p vod systems,

    Y . Zhang, C. Gao, Y . Guo, K. Bian, X. Jin, Z. Yang, L. Song, J. Cheng, H. Tuo, and X. Li, “Proactive video push for optimizing bandwidth consumption in hybrid cdn-p2p vod systems,” inIEEE INFOCOM 2018 - IEEE Conference on Computer Communications, 2018, pp. 2555–2563

    2018

  11. [19]

    Intelligent edge-assisted crowdcast with deep reinforcement learning for personalized qoe,

    F. Wang, C. Zhang, F. wang, J. Liu, Y . Zhu, H. Pang, and L. Sun, “Intelligent edge-assisted crowdcast with deep reinforcement learning for personalized qoe,” inIEEE INFOCOM 2019 - IEEE Conference on Computer Communications, 2019, pp. 910–918

    2019

  12. [20]

    Globally distributed content delivery,

    J. Dilley, B. Maggs, J. Parikh, H. Prokop, R. Sitaraman, and B. Weihl, “Globally distributed content delivery,”IEEE Internet Computing, vol. 6, no. 5, pp. 50–58, 2002

    2002

  13. [21]

    Leveraging qoe heterogenity for large-scale livecaset scheduling,

    R.-X. Zhang, M. Ma, T. Huang, H. Li, J. Liu, and L. Sun, “Leveraging qoe heterogenity for large-scale livecaset scheduling,” inProceedings of the 28th ACM International Conference on Multimedia, 2020, pp. 3678–3686

    2020

  14. [22]

    Deepcast: Towards personalized qoe for edge-assisted crowdcast with deep reinforcement learning,

    F. Wang, C. Zhang, F. Wang, J. Liu, Y . Zhu, H. Pang, and L. Sun, “Deepcast: Towards personalized qoe for edge-assisted crowdcast with deep reinforcement learning,”IEEE/ACM Transactions on Networking, vol. 28, no. 3, pp. 1255–1268, 2020

    2020

  15. [23]

    Livesmart: A qos-guaranteed cost-minimum framework of viewer scheduling for crowdsourced live streaming,

    R.-X. Zhang, M. Ma, T. Huang, H. Pang, X. Yao, C. Wu, J. Liu, and L. Sun, “Livesmart: A qos-guaranteed cost-minimum framework of viewer scheduling for crowdsourced live streaming,” inProceedings of the 27th ACM International Conference on Multimedia, 2019, pp. 420–428

    2019

  16. [24]

    A recurrent latent variable model for sequential data,

    J. Chung, K. Kastner, L. Dinh, K. Goel, A. C. Courville, and Y . Bengio, “A recurrent latent variable model for sequential data,”Advances in neural information processing systems, vol. 28, 2015

    2015

  17. [25]

    World cup streaming suffers major problems in australia

    Optus-sports, “World cup streaming suffers major problems in australia.” 2018. [Online]. Available: https://www.dacast.com/blog/blog- live-streaming-cdn-pricing/

    2018

  18. [26]

    Intelligent edge learning for personalized crowdsourced livecast: Challenges, opportuni- ties, and solutions,

    F. Wang, J. Liu, C. Zhang, L. Sun, and K. Hwang, “Intelligent edge learning for personalized crowdsourced livecast: Challenges, opportuni- ties, and solutions,”IEEE Network, vol. 35, no. 1, pp. 170–176, 2021

    2021

  19. [27]

    A large-scale holistic measurement of crowdsourced edge cloud platform,

    Y . Feng, S. Shen, M. Xu, C. Zhang, X. Wang, X. Wang, W. Wang, and V . C. Leung, “A large-scale holistic measurement of crowdsourced edge cloud platform,”World Wide Web, vol. 26, no. 5, pp. 3561–3584, 2023

    2023

  20. [28]

    Auto-encoding variational bayes,

    D. P. Kingma and M. Welling, “Auto-encoding variational bayes,”arXiv preprint arXiv:1312.6114, 2013

    2013 arXiv

  21. [29]

    Time-series anomaly detection service at microsoft,

    H. Ren, B. Xu, Y . Wang, C. Yi, C. Huang, X. Kou, T. Xing, M. Yang, J. Tong, and Q. Zhang, “Time-series anomaly detection service at microsoft,” inProceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, 2019

    2019

  22. [30]

    Electric vehicle fleet size and trip pricing for one-way carsharing services considering vehicle relocation and per- sonnel assignment,

    M. Xu, Q. Meng, and Z. Liu, “Electric vehicle fleet size and trip pricing for one-way carsharing services considering vehicle relocation and per- sonnel assignment,”Transportation Research Part B: Methodological, vol. 111, pp. 60–82, 2018

    2018

  23. [31]

    Variational inference for dirichlet process mixtures,

    D. M. Blei and M. I. Jordan, “Variational inference for dirichlet process mixtures,” 2006

    2006

  24. [32]

    Graph wavenet for deep spatial-temporal graph modeling,

    Z. Wu, S. Pan, G. Long, J. Jiang, and C. Zhang, “Graph wavenet for deep spatial-temporal graph modeling,” inProceedings of the 28th International Joint Conference on Artificial Intelligence, 2019

    2019

  25. [33]

    Robust anomaly detection for multivariate time series through stochastic recurrent neural network,

    Y . Su, Y . Zhao, C. Niu, R. Liu, W. Sun, and D. Pei, “Robust anomaly detection for multivariate time series through stochastic recurrent neural network,” inProceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, 2019

    2019

  26. [34]

    Seer: Proactive revenue-aware scheduling for live streaming services in crowdsourced cloud-edge platforms,

    S. Huang, Z. Wang, Z. Zhang, H. Zhang, X. Wang, and W. Wang, “Seer: Proactive revenue-aware scheduling for live streaming services in crowdsourced cloud-edge platforms,”IEEE INFOCOM 2024 - IEEE Conference on Computer Communications, 2024

    2024

  27. [35]

    Iot scheduling for higher throughput and lower transmission power,

    H. Rajab, T. Cinkler, and T. Bouguera, “Iot scheduling for higher throughput and lower transmission power,”Wireless Networks, vol. 27, pp. 1701–1714, 2021

    2021

  28. [36]

    Towards optimizing task scheduling process in cloud environment,

    Y . Shi, K. Suo, J. Hodge, D. P. Mohandoss, and S. Kemp, “Towards optimizing task scheduling process in cloud environment,” in2021 IEEE 11th annual computing and communication workshop and conference (CCWC). IEEE, 2021, pp. 0081–0087

    2021

  29. [37]

    A scheduling algorithm to maximize storm throughput in heterogeneous cluster,

    H. Nasiri, S. Nasehi, A. Divband, and M. Goudarzi, “A scheduling algorithm to maximize storm throughput in heterogeneous cluster,” Journal of Big Data, vol. 10, no. 1, p. 103, 2023

    2023

  30. [38]

    Machine learning methods for reliable resource provisioning in edge-cloud computing: A survey,

    T. L. Duc, R. G. Leiva, P. Casari, and P.-O. ¨Ostberg, “Machine learning methods for reliable resource provisioning in edge-cloud computing: A survey,”ACM Computing Surveys (CSUR), 2019

    2019

  31. [39]

    From cloud to edge: a first look at public edge platforms,

    M. Xu, Z. Fu, X. Ma, L. Zhang, Y . Li, F. Qian, S. Wang, K. Li, J. Yang, and X. Liu, “From cloud to edge: a first look at public edge platforms,” in Proceedings of the 21st ACM Internet Measurement Conference, 2021, pp. 37–53

    2021

  32. [40]

    A branch-and-cut algorithm for the dial-a-ride problem,

    J.-F. Cordeau, “A branch-and-cut algorithm for the dial-a-ride problem,” Operations Research, vol. 54, no. 3, pp. 573–586, 2006

    2006

  33. [41]

    A universal transcoding and transmission method for livecast with networked multi-agent reinforcement learning,

    X. Chen, C. Xu, M. Wang, Z. Wu, S. Yang, L. Zhong, and G.-M. Muntean, “A universal transcoding and transmission method for livecast with networked multi-agent reinforcement learning,” inIEEE INFOCOM 2021 - IEEE Conference on Computer Communications, 2021, pp. 1–10

    2021

  34. [42]

    Optimal control for generalized network- flow problems,

    A. Sinha and E. Modiano, “Optimal control for generalized network- flow problems,”IEEE/ACM Transactions on Networking, vol. 26, no. 1, pp. 506–519, 2018

    2018

  35. [43]

    Rldish: Edge-assisted qoe optimization of http live streaming with reinforcement learning,

    H. Wang, K. Wu, J. Wang, and G. Tang, “Rldish: Edge-assisted qoe optimization of http live streaming with reinforcement learning,” inIEEE INFOCOM 2020 - IEEE Conference on Computer Communications, 2020, pp. 706–715

    2020

  36. [44]

    One for all: Unified workload prediction for dynamic multi-tenant edge cloud platforms,

    S. Huang, Z. Wang, H. Zhang, X. Wang, C. Zhang, and W. Wang, “One for all: Unified workload prediction for dynamic multi-tenant edge cloud platforms,” inProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2023, pp. 788–797

    2023

  37. [45]

    Spatio-temporal-social multi-feature-based fine-grained hot spots prediction for content delivery services in 5g era,

    S. Huang, H. Zhang, X. Wang, M. Chen, J. Li, and V . C. M. Le- ung, “Spatio-temporal-social multi-feature-based fine-grained hot spots prediction for content delivery services in 5g era,” inProceedings of the 30th ACM International Conference on Information & Knowledge Managem...

    2021

  38. [46]

    Com- boptnet: Fit the right np-hard problem by learning integer programming constraints,

    A. Paulus, M. Rolinek, V . Musil, B. Amos, and G. Martius, “Com- boptnet: Fit the right np-hard problem by learning integer programming constraints,” inProceedings of the 38th International Conference on Machine Learning, ser. Proceedings of Machine Learning Research. PMLR, 20...

    2021

  39. [47]

    Lalisa: Adaptive bitrate ladder optimization in http-based adaptive live streaming,

    F. Tashtarian, A. Bentaleb, H. Amirpour, B. Taraghi, C. Timmerer, H. Hellwagner, and R. Zimmermann, “Lalisa: Adaptive bitrate ladder optimization in http-based adaptive live streaming,” inNOMS 2023-2023 IEEE/IFIP Network Operations and Management Symposium, 2023, pp. 1–9

    2023

  40. [48]

    A systematic review of service level management in the cloud,

    F. Faniyi and R. Bahsoon, “A systematic review of service level management in the cloud,”ACM Comput. Surv., vol. 48, no. 3, 2015

    2015

  41. [49]

    Qoe management of multimedia stream- ing services in future networks: A tutorial and survey,

    A. A. Barakabitze, N. Barman, A. Ahmad, S. Zadtootaghaj, L. Sun, M. G. Martini, and L. Atzori, “Qoe management of multimedia stream- ing services in future networks: A tutorial and survey,”IEEE Commu- nications Surveys & Tutorials, vol. 22, no. 1, pp. 526–565, 2019

    2019

  42. [50]

    4d-map: Multipath adaptive packet scheduling for live streaming over quic,

    C.-X. Song, B. Han, and J.-S. Su, “4d-map: Multipath adaptive packet scheduling for live streaming over quic,”Journal of Computer Science and Technology, vol. 39, no. 1, pp. 159–176, 2024

    2024

  43. [51]

    Towards optimal low- latency live video streaming,

    L. Sun, T. Zong, S. Wang, Y . Liu, and Y . Wang, “Towards optimal low- latency live video streaming,”IEEE/ACM transactions on networking, vol. 29, no. 5, pp. 2327–2338, 2021

    2021

  44. [52]

    Tclivi: Transmission control in live video streaming based on deep reinforcement learning,

    L. Cui, D. Su, S. Yang, Z. Wang, and Z. Ming, “Tclivi: Transmission control in live video streaming based on deep reinforcement learning,” IEEE Transactions on Multimedia, vol. 23, pp. 651–663, 2020

    2020

  45. [53]

    Greenabr: Energy-aware adaptive bitrate streaming with deep reinforcement learning,

    B. O. Turkkan, T. Dai, A. Raman, T. Kosar, C. Chen, M. F. Bulut, J. Zola, and D. Sow, “Greenabr: Energy-aware adaptive bitrate streaming with deep reinforcement learning,” inProceedings of the 13th ACM Multimedia Systems Conference, 2022, pp. 150–163

    2022

  46. [54]

    Adaptive qos-aware multipath congestion control for live streaming,

    X. Ji, B. Han, C. Xu, C. Song, and J. Su, “Adaptive qos-aware multipath congestion control for live streaming,”Computer Networks, vol. 220, p. 109470, 2023

    2023

  47. [55]

    Evaluation of machine learn- ing algorithms for anomaly detection,

    N. Elmrabit, F. Zhou, F. Li, and H. Zhou, “Evaluation of machine learn- ing algorithms for anomaly detection,” in2020 international conference on cyber security and protection of digital services (cyber security). IEEE, 2020, pp. 1–8

    2020

  48. [56]

    Cur- coedge: Curiosity-driven collaborative request scheduling in edge-cloud systems,

    Y . Zhao, C. Qiu, X. Shi, X. Wang, D. Niyato, and V . C. Leung, “Cur- coedge: Curiosity-driven collaborative request scheduling in edge-cloud systems,” inIEEE INFOCOM 2024-IEEE Conference on Computer Communications. IEEE, 2024, pp. 411–420

    2024

  49. [57]

    Fine-grained spatio-temporal distribution prediction of mobile content delivery in 5g ultra-dense networks,

    S. Huang, H. Zhang, X. Wang, M. Chen, J. Li, and V . C. Leung, “Fine-grained spatio-temporal distribution prediction of mobile content delivery in 5g ultra-dense networks,”IEEE Transactions on Mobile Computing, vol. 23, no. 1, pp. 469–482, 2022

    2022

  50. [58]

    Qarc: Video quality aware rate control for real-time video streaming based on deep reinforcement learning,

    T. Huang, R.-X. Zhang, C. Zhou, and L. Sun, “Qarc: Video quality aware rate control for real-time video streaming based on deep reinforcement learning,” inProceedings of the 26th ACM international conference on Multimedia, 2018, pp. 1208–1216

    2018

  51. [59]

    Ppo-abr: Proximal policy optimization based deep reinforcement learning for adaptive bitrate streaming,

    M. Naresh, P. Saxena, and M. Gupta, “Ppo-abr: Proximal policy optimization based deep reinforcement learning for adaptive bitrate streaming,” in2023 International Wireless Communications and Mobile Computing (IWCMC). IEEE, 2023, pp. 199–204

    2023

  52. [60]

    Deeplive: Qoe opti- mization for live video streaming through deep reinforcement learning,

    Z. Tian, L. Zhao, L. Nie, P. Chen, and S. Chen, “Deeplive: Qoe opti- mization for live video streaming through deep reinforcement learning,” in2019 IEEE 25th international conference on parallel and distributed systems (ICPADS). IEEE, 2019, pp. 827–831

    2019

  53. [61]

    Online anomaly detection over live social video streaming,

    C. He, X. Zhou, C. Wang, I. Gondal, J. Shao, and X. Yi, “Online anomaly detection over live social video streaming,”arXiv preprint arXiv:2401.08615, 2023

    2023 arXiv

  54. [62]

    Digital twin- based job shop anomaly detection and dynamic scheduling,

    Y . Li, Z. Tao, L. Wang, B. Du, J. Guo, and S. Pang, “Digital twin- based job shop anomaly detection and dynamic scheduling,”Robotics and Computer-Integrated Manufacturing, vol. 79, p. 102443, 2023

    2023

  55. [63]

    Adaptive anomaly detection for iot data in hierarchical edge computing,

    M. V . Ngo, H. Chaouchi, T. Luo, and T. Q. Quek, “Adaptive anomaly detection for iot data in hierarchical edge computing,”arXiv preprint arXiv:2001.03314, 2020

    2001 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.