Pith. sign in

REVIEW 3 major objections 5 minor 94 references

ConfigTron: Tackling network diversity with heterogeneous configurations

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read ConfigTron shows that per-connection TCP and HTTP configuration tuning cuts median page load times by up to 19% in simulation and 8–10% in a live deployment.

desk verdict A solid systems paper with a real live-deployment result; the headline 19% comes from a simulator that reads its own testbed-built oracle, so treat the sim numbers as upper bounds, not measured gains. read the letter →

arxiv 1908.04518 v1 pith:VJIQ3E3U submitted 2019-08-13 cs.NI cs.LG

Usama Naseer , Theophilus Benson This is my paper · ORCID
classification cs.NIcs.LG
keywords webperformancepageloadtimenetworkconfigurationtuningcontextualmulti-armedbanditcongestioncontrolselectionHTTP/2CDNedgeserversclasses
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 the one-size-fits-all configuration of web-serving stacks is suboptimal because end-user networks, devices, and websites differ enough that no single protocol and parameter choice serves everyone well. To exploit this diversity, it presents ConfigTron, a framework that lets CDN edge servers reconfigure TCP and HTTP knobs per connection, driven by a contextual multi-armed bandit that clusters similar clients and learns from all connections. The paper's core evidence is that choosing configurations per network class and website improves median page load time by up to 19% (about 500–750 ms) in trace-driven simulation and by 8–10% in a live deployment, with larger tail-latency gains. A sympathetic reader would take the paper as establishing that the gains from heterogeneous, dynamically learned configurations are real and worth the engineering cost, especially for users on high-RTT, high-loss networks.

What carries the argument

The load-bearing mechanism is the contextual multi-armed bandit ensemble: a Gaussian-process arm performs directed exploration to quickly find a good configuration, an epsilon-greedy arm randomly re-samples old points to handle non-stationary conditions and non-Gaussian noise, and a decision-tree arm exploits accumulated data to predict the best configuration. Clients are grouped into network classes via k-means clustering, and the learning process is bootstrapped with Latin Hypercube Sampling. The whole pipeline is trained and evaluated against the PLT-Tensor, a precomputed lookup table mapping each {network condition, configuration, website} triple to a measured page load time; the simulator computes a client's PLT by looking up this tensor, and the learned decision tree is converted into key-value rules that edge servers consult in real time.

What would settle it

Measure real page load times for matched {network condition, configuration, website} triples on production edge servers and compare them with the PLT-Tensor's predictions; if the tensor is systematically off, or if a live A/B test against defaults fails to reproduce the 8–10% median improvement, the central claim collapses.

Watch

Extended reading notes

Core claim

The central claim is that a one-size-fits-all configuration of the web-serving stack leaves substantial performance on the table, and that a data-driven system can recover it. ConfigTron learns, for each group of similar clients and each website, which combination of congestion-control algorithm, initial congestion window, TCP settings, and HTTP version minimizes page load time. It does this online with a contextual multi-armed bandit that clusters clients into network classes, explores with a Gaussian-process arm and an epsilon-greedy arm, and exploits with a decision tree. The paper reports median page-load-time reductions of up to 19% (about 750 ms) in trace-driven simulation and 8–10% in a live deployment spread across roughly 100K page loads, with larger gains in the tail and in high-RTT, high-loss networks representative of developing regions. In the authors' telling, even modest median gains translate into meaningful revenue for content providers.

Load-bearing premise

ConfigTron's simulation gains rest on the PLT-Tensor, a table of page load times built once in a testbed, correctly predicting what real browsers on real devices would experience across 16 million simulated sessions.

Editorial extensions

If this is right

  • CDNs can deploy per-connection configuration without client-side changes: the Configuration Manager pushes decision-tree rules as key-value maps, and the edge agent applies them at connection setup, avoiding startup delay.
  • The largest wins appear on high-RTT, high-loss, low-bandwidth paths typical of developing regions, where defaults tuned for developed-region networks are most mismatched.
  • ConfigTron's top configurations remain largely fair to Cubic traffic (Jain index above 0.95 in more than half of tested cases), and removing BBR from the candidate set pushes fairness above 0.95 for more than 90% of cases.
  • The same learning loop extends naturally to newer transports like QUIC and BBR, which expose more configuration parameters and therefore strengthen the case for automated, per-connection tuning.

Reading between the lines

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

  • The simulated 19% headline is probably an upper bound: because the simulator looks up PLT from a testbed-built tensor and omits device heterogeneity, browser rendering, and TCP slow-start-after-idle, the live 8–10% median figure is the more reliable estimate for real deployments.
  • The paper's own Pantheon analysis, showing that no congestion control is optimal for more than 25% of networks, implies that protocol innovation alone will not close the tuning gap; selecting among protocols is itself a first-class optimization axis.
  • A testable extension would be to replace the PLT-Tensor lookup with a lightweight online model trained on real production traffic, then check whether ConfigTron's convergence behavior and gains survive outside the testbed's network-condition distribution.
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

3 major / 5 minor

Summary. The paper argues that the one-size-fits-all configuration of web-serving protocol stacks is suboptimal and presents ConfigTron, a data-driven framework that selects per-connection TCP and HTTP configurations using a contextual multi-armed bandit ensemble with a decision-tree exploitation arm. The system includes a kernel-module-based ConfigTron-API for fine-grained reconfiguration, a centralized Configuration Manager, and a push-based Configuration Agent. The authors evaluate ConfigTron in two ways: large-scale trace-driven simulations using a precomputed PLT-Tensor that maps network conditions, configurations, and websites to page load times, and a live deployment on AWS with globally distributed Speedchecker clients. They report median PLT improvements of up to 19% (Pantheon), 16% (MAWI), 11.2% (FCC), and 10.1% (CAIDA) in simulation, and 8-10% median and 43% 95th-percentile improvements in the live deployment. The paper also includes an empirical study of configuration heterogeneity across CSPs and regions, a fairness analysis, and a discussion of design choices.

Significance. If the central claims hold, ConfigTron is a valuable contribution to network-stack configuration tuning: it provides a concrete architecture for fine-grained per-connection control, a learning algorithm combining guided and stochastic exploration, and live-deployment evidence that tuning can improve end-user page load times. The live deployment, in particular, is a meaningful validation because it measures real page loads against a control group, and the oracle baseline used in simulation is conservatively given perfect network knowledge. The empirical study of configuration heterogeneity (Table 2) is also a useful measurement contribution. However, the headline simulation numbers are undermined by a circular evaluation methodology (the PLT-Tensor is both the training signal and the ground truth), and the paper lacks sensitivity analysis for several key hyperparameters. The qualitative conclusion that heterogeneous tuning helps survives, but the quantitative strength of the claim is not yet established.

major comments (3)
  1. [6.1 (Simulator and Limitations) and 6.2 (Effectiveness)]
  2. [4 (Algorithm) and 5 (Prototype)]
  3. [6.2 (Effectiveness) and Figures 4-9]
minor comments (5)
  1. [6.4.3 (Microbenchmarks)]
  2. [6.6 (Critical Features and Parameters)]
  3. [2.2 (Implications of Configuration Tuning)]
  4. [6.7 (Live Deployment)]
  5. [4.1.1 (Prediction with Machine Learning)]

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ConfigTron's learning algorithm is evaluated against an externally measured PLT-Tensor with a live-deployment check, and no prediction reduces to its fitted inputs by construction.

full rationale

The paper's derivation chain is not circular. The PLT-Tensor is built from physical testbed measurements: for each {network condition, configuration, website} triple, webpages are loaded with a real browser and the median PLT is recorded (§6.1, 'Optimal Configurations'). ConfigTron's contextual bandit and decision tree do not receive the tensor's optimal labels; they must explore configurations and receive PLT rewards through the simulator's lookup. The simulation then reports PLT improvements by reading the same tensor, so the simulated gains are upper bounds on the fidelity of the testbed's PLT model rather than independent protocol-stack validation. This is a benchmark limitation, not a constructed circularity: the predicted quantity (which configuration is best) is not definitionally equal to the fitted input (the tensor), and the learner has no direct access to the argmin. The live deployment in §6.7 provides an external, tensor-independent evaluation showing 8-10% median gains, which corroborates the qualitative claim. The only self-citations are tooling or background references (e.g., [61] InspectorGadget for active probing, [2] for mobile-device measurements) and are not load-bearing for the central claim. No equation is shown to reduce to its own inputs, no fitted parameter is renamed as a prediction, and no uniqueness theorem or prior result by the same authors is invoked to force the design. Therefore the paper is self-contained with respect to its claimed derivation, and the circularity score is 0.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

ConfigTron's quantitative claims rest on several domain assumptions that are cited from prior work rather than demonstrated here: the non-stationarity of networks, the equivalence of similar network conditions for configuration purposes, and the fidelity of the testbed-derived PLT-Tensor. The hyperparameters controlling the bandit ensemble are chosen by hand and not swept in the paper, so the reported gains are conditional on those choices.

free parameters (7)
  • EI threshold = 5%
    Gaussian process bandit transitions to epsilon exploration when Expected Improvement exceeds 5%; chosen by the authors in Section 5, not derived.
  • GP init_sample = 4
    Number of initial samples for Gaussian process bootstrapping; set in Section 5.
  • GP min_sample_tested = 7
    Minimum samples tested before terminating GP exploration; set in Section 5.
  • Decision tree min samples leaf = 80
    Minimum number of leaf node samples in CART; chosen in Section 5.
  • Decision tree min samples split = 2
    Minimum samples for a split; chosen in Section 5.
  • Number of network classes k = not specified
    Determined by kmeans clustering with the criterion that performance spread within a class is bounded by one standard deviation from the mean (Section 4.2); the actual k is not reported.
  • Epsilon degree of randomness = not specified
    Operator-bounded parameter trading exploration speed against QoE impact (Section 4.1.2); no value is given.
assumptions (6)
  • domain assumption Users with very similar network conditions will require identical optimal configurations.
    Section 4.2 states this assumption, citing [44,60]; if false, clustering clients into network classes and assigning one configuration per class would be suboptimal.
  • domain assumption Network conditions are non-stationary and change every few minutes.
    Section 1.2 and 4 list non-stationarity as a core design driver, citing [10,44,55,84,93]; the need for continuous online learning depends on this.
  • domain assumption The PLT-Tensor built from testbed emulation is a faithful model of f(c,n,w) for all simulated sessions.
    Section 6.1 simulates page loads by looking up the tensor; the simulator does not run the real protocol stack, so all simulated improvement numbers inherit the tensor's fidelity.
  • domain assumption Testbed-emulated network conditions drawn from CAIDA, MAWI, FCC, and Pantheon are representative of the real Internet conditions faced by CDN users.
    Section 6.1 uses these four datasets to both construct the tensor and generate client sessions; generalization to other conditions is assumed.
  • domain assumption Page load time is the appropriate performance metric and captures end-user QoE.
    Section 4 formulates the objective with PLT and notes other metrics can replace it; the evaluation only uses PLT, so results may not extend to SpeedIndex or other QoE metrics.
  • domain assumption Latin Hypercube Sampling improves GP bootstrapping relative to random or domain-specific sampling.
    Section 4.3 adopts LHC based on [57] and claims 2-3x reduction in optimization steps; not independently verified in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ConfigTron: Tackling network diversity with heterogeneous configurations." pith.science (2026). https://pith.science/paper/VJIQ3E3U

@misc{pith2026190804518,
  author       = {Pith},
  title        = {Pith review of: ConfigTron: Tackling network diversity with heterogeneous configurations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VJIQ3E3U}},
  note         = {Machine review of arXiv:1908.04518}
}
read the original abstract

The web serving protocol stack is constantly changing and evolving to tackle technological shifts in networking infrastructure and website complexity. As a result of this evolution, the web serving stack includes a plethora of protocols and configuration parameters that enable the web serving stack to address a variety of realistic network conditions. Yet, today, most content providers have adopted a "one-size-fits-all" approach to configuring the networking stack of their user facing web servers (or at best employ moderate tuning), despite the significant diversity in end-user networks and devices. In this paper, we revisit this problem and ask a more fundamental question: Are there benefits to tuning the network stack? If so, what system design choices and algorithmic ensembles are required to enable modern content provider to dynamically and flexibly tune their protocol stacks. We demonstrate through substantial empirical evidence that this "one-size-fits-all" approach results in sub-optimal performance and argue for a novel framework that extends existing CDN architectures to provide programmatic control over the configuration options of the CDN serving stack. We designed ConfigTron a data-driven framework that leverages data from all connections to identify their network characteristics and learn the optimal configuration parameters to improve end-user performance. ConfigTron uses contextual multi-arm bandit-based learning algorithm to find optimal configurations in minimal time, enabling a content providers to systematically explore heterogeneous configurations while improving end-user page load time by as much as 19% (upto 750ms) on median.

Figures

Figures reproduced from arXiv: 1908.04518 by the authors.

Figure 1
Figure 1. PLT comparison improve quality of experience (QoE), the networking com￾munity is constantly developing new protocols and configu￾ration suggestions for user-facing web servers (AKA, edge servers), e.g., PCC [24], Vivace [25], and BBR [20] at the congestion layer (L4) and QUIC [38], SPDY [58] and HTTP2 at the application layer (L7). The optimal choice of protocol and parameters is con￾tingent on the network infrastru… view at source ↗
Figure 3
Figure 3. Learning [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. CAIDA traces [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figures from the paper (4 more)
Figure 6
Figure 6. Figure 6: Convergence 9 [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Jain’s Fairness ICWs (eg. 30 as ICW). In order to evaluate fairness, we sim￾ulate an end-user’s last-mile connection in our testbed. We test multiple representative network conditions (3G, 4G, Ca￾ble, DSL etc) with shallow buffers. In each network, we start 5 backgroun…
Figure 8
Figure 8. Figure 8: Critical features [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 10
Figure 10. Figure 10: Live deployment conducted pageloads throughout the day, resulting in 100K pageloads in 10 days. We observed that ConfigTron improves the web perfor￾mance by 8-10% in the median and 43% in the 95th percentile case. For the highest improvements scenarios, the optimal co…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

94 extracted references · 76 canonical work pages

  1. [1]

    P., and MacKay, D

    Adams, R. P., and MacKay, D. J. Bayesian online changepoint detec- tion. arXiv preprint arXiv:0710.3742 (2007)

    arXiv 2007

  2. [2]

    L., Qazi, Z

    Ahmad, S., Haamid, A. L., Qazi, Z. A., Zhou, Z., Benson, T., and Qazi, I. A. A view from the other side: Understanding mobile phone characteristics in the developing world. InProceedings of the 2016 ACM on Internet Measurement Conference (2016), ACM, pp. 319–325

    2016

  3. [3]

    Tcp optimizations

    Akamai. Tcp optimizations. https://developer.akamai.com/legacy/ learn/Optimization/TCP_Optimizations.html

  4. [4]

    S., Govindan, R., Rao, S., Chen, J., Katz-Bassett, E., Ribeiro, B., Zhan, J., and Zhang, H

    Akhtar, Z., Nam, Y. S., Govindan, R., Rao, S., Chen, J., Katz-Bassett, E., Ribeiro, B., Zhan, J., and Zhang, H. Oboe: auto-tuning video abr algorithms to network conditions. InProceedings of the 2018 Conference of the ACM Special Interest Group on Data Communication (2018), ACM, pp. 44–58

    2018

  5. [5]

    Overclock- ing the yahoo!: Cdn for faster web page loads

    Al-Fares, M., Elmeleegy, K., Reed, B., and Gashinsky, I. Overclock- ing the yahoo!: Cdn for faster web page loads. In Proceedings of the 2011 ACM SIGCOMM conference on Internet measurement conference (2011), ACM, pp. 569–584

    2011

  6. [6]

    H., Chen, J., Venkataraman, S., Yu, M., and Zhang, M

    Alipourfard, O., Liu, H. H., Chen, J., Venkataraman, S., Yu, M., and Zhang, M. Cherrypick: Adaptively unearthing the best cloud configurations for big data analytics. In NSDI (2017), pp. 469–482

    2017

  7. [7]

    Apache benchmark

    Apache. Apache benchmark. https://httpd.apache.org/docs/2.4/ programs/ab.html

  8. [8]

    Archive, M. W. G. T. Packet traces from wide backbone 12/1/17 to 12/7/17. http://mawi.wide.ad.jp/mawi/

Show all 94 references
  1. [9]

    S., and Seshan, S

    Balakrishnan, H., Rahul, H. S., and Seshan, S. An integrated con- gestion management architecture for internet hosts. ACM SIGCOMM Computer Communication Review 29 , 4 (1999), 175–187

    1999

  2. [10]

    Balakrishnan, H., Stemm, M., Seshan, S., and Katz, R. H. Analyz- ing stability in wide-area network performance. ACM SIGMETRICS Performance Evaluation Review 25 , 1 (1997), 2–12

    1997

  3. [11]

    Towards automatic parameter tuning of stream processing systems

    Bilal, M., and Canini, M. Towards automatic parameter tuning of stream processing systems. In Proceedings of the 2017 Symposium on Cloud Computing (New York, NY, USA, 2017), SoCC ’17, ACM, pp. 189– 200

    2017

  4. [12]

    Tcp-bpf: Programmatically tuning tcp behavior through bpf

    Brakmo, L. Tcp-bpf: Programmatically tuning tcp behavior through bpf. NetDev 2.2 (2017)

    2017

  5. [13]

    G., Wong, G

    Bridges, P. G., Wong, G. T., Hiltunen, M., Schlichting, R. D., and Barrick, M. J. A configurable and extensible transport protocol. IEEE/ACM Transactions on Networking 15 , 6 (2007), 1254–1265

    2007

  6. [14]

    M., and De Freitas, N

    Brochu, E., Cora, V. M., and De Freitas, N. A tutorial on bayesian optimization of expensive cost functions, with application to active user modeling and hierarchical reinforcement learning. arXiv preprint arXiv:1012.2599 (2010)

    2010 arXiv

  7. [15]

    Analysis of RouteViews BGP data: pol- icy atoms

    Broido, A., and claffy, k. Analysis of RouteViews BGP data: pol- icy atoms. In Network Resource Data Management Workshop (Santa Barbara, CA, May 2001)

    2001

  8. [16]

    Speed matters for google web search, 2009

    Brutlag, J. Speed matters for google web search, 2009

    2009

  9. [17]

    V., and Sekar, V

    Butkiewicz, M., Madhyastha, H. V., and Sekar, V. Understanding website complexity: measurements, metrics, and implications. In Pro- ceedings of the 2011 ACM SIGCOMM conference on Internet measurement conference (2011), ACM, pp. 313–328

    2011

  10. [18]

    V., and Sekar, V

    Butkiewicz, M., W ang, D., Wu, Z., Madhyastha, H. V., and Sekar, V. Klotski: Reprioritizing web content to improve user experience on mobile devices. In NSDI (2015), vol. 1, pp. 2–3

    2015

  11. [19]

    The caida ucsd anonymized internet traces 2016 dataset

    CAIDA. The caida ucsd anonymized internet traces 2016 dataset. http://www.caida.org/data/passive/passive_2016_dataset.xml

    2016

  12. [20]

    S., Yeganeh, S

    Cardwell, N., Cheng, Y., Gunn, C. S., Yeganeh, S. H., et al. Bbr: congestion-based congestion control. Communications of the ACM 60 , 2 (2017), 58–66

    2017

  13. [21]

    T., Phan, L

    Chen, A., Sriraman, A., V aidya, T., Zhang, Y., Haeberlen, A., Loo, 13 , Usama Naseer Theophilus Benson Brown University B. T., Phan, L. T. X., Sherr, M., Shields, C., and Zhou, W. Dispersing asymmetric ddos attacks with splitstack. In HotNets (2016), pp. 197– 203

    2016

  14. [22]

    Dean, J., and Barroso, L. A. The tail at scale. Communications of the ACM 56 (2013), 74–80

    2013

  15. [23]

    Dynamo: amazon’s highly available key-value store

    DeCandia, G., Hastorun, D., Jampani, M., Kakulapati, G., Laksh- man, A., Pilchin, A., Sivasubramanian, S., Vosshall, P., and Vo- gels, W. Dynamo: amazon’s highly available key-value store. In ACM SIGOPS operating systems review (2007), vol. 41, ACM, pp. 205–220

    2007

  16. [24]

    B., and Schapira, M

    Dong, M., Li, Q., Zarchy, D., Godfrey, P. B., and Schapira, M. Pcc: Re-architecting congestion control for consistent high performance. In NSDI (2015), vol. 1, p. 2

    2015

  17. [25]

    Pcc vivace: Online-learning congestion con- trol

    Dong, M., Meng, T., Zarchy, D., Arslan, E., Gilad, Y., Godfrey, B., and Schapira, M. Pcc vivace: Online-learning congestion con- trol. In 15th{USENIX} Symposium on Networked Systems Design and Implementation ({NSDI} 18). USENIX Association (2018)

    2018

  18. [26]

    Tuning database configuration parameters with ituned

    Duan, S., Thummala, V., and Babu, S. Tuning database configuration parameters with ituned. Proceedings of the VLDB Endowment 2, 1 (2009), 1246–1257

    2009

  19. [27]

    An argument for increasing tcp’s initial congestion window

    Dukkipati, N., Refice, T., Cheng, Y., Chu, J., Herbert, T., Agarwal, A., Jain, A., and Sutin, N. An argument for increasing tcp’s initial congestion window. Computer Communication Review 40 , 3 (2010), 26–33

    2010

  20. [28]

    The lwip tcp/ip stack

    Dunkels, A., et al. The lwip tcp/ip stack. lwIP–A LightWeight TCP/IP Stack (2004)

    2004

  21. [29]

    A generic hash table

    Edge, J. A generic hash table

  22. [30]

    Erman, J., Gopalakrishnan, V., Jana, R., and Ramakrishnan, K. K. Towards a spdyâĂŹier mobile web? IEEE/ACM Transactions on Net- working 23, 6 (2015), 2010–2023

    2015

  23. [31]

    Aquila (internet deployment by drone)

    Facebook. Aquila (internet deployment by drone). https://code. facebook.com/posts/268598690180189

  24. [32]

    Network connection class

    Facebook. Network connection class. https://github.com/facebook/ network-connection-class

  25. [33]

    Advanced tcp optimizations

    Fastly. Advanced tcp optimizations. https://docs.fastly.com/guides/ live-streaming/configuration-guidelines-for-live-streaming

  26. [34]

    Measuring fixed broadband report - 2016

    FCC. Measuring fixed broadband report - 2016. https://www. fcc.gov/reports-research/reports/measuring-broadband-america/ measuring-fixed-broadband-report-2016

    2016

  27. [35]

    Open vSwitch

    foundation, L. Open vSwitch

  28. [36]

    R., Kusner, M

    Gardner, J. R., Kusner, M. J., Xu, Z. E., Weinberger, K. Q., and Cunningham, J. P. Bayesian optimization with inequality constraints. In ICML (2014), pp. 937–945

    2014

  29. [37]

    H., and Ganjali, Y

    Ghobadi, M., Yeganeh, S. H., and Ganjali, Y. Rethinking end-to-end congestion control in software-defined networks. In Proceedings of the 11th ACM Workshop on Hot Topics in networks (2012), ACM, pp. 61–66

    2012

  30. [38]

    Quic, a multiplexed stream transport over udp

    Google. Quic, a multiplexed stream transport over udp. https://www. chromium.org/quic

  31. [39]

    Guide, T. L. K. M. P. Character Device Drivers

  32. [40]

    B., and Babu, S

    Herodotou, H., Lim, H., Luo, G., Borisov, N., Dong, L., Cetin, F. B., and Babu, S. Starfish: A self-tuning system for big data analytics. In Cidr (2011), vol. 11, pp. 261–272

    2011

  33. [41]

    Experimental evaluation of bbr congestion control

    Hock, M., Bless, R., and Zitterbart, M. Experimental evaluation of bbr congestion control. In Network Protocols (ICNP), 2017 IEEE 25th International Conference on (2017), IEEE, pp. 1–10

    2017

  34. [42]

    H., Godfrey, P., Schapira, M., and Tamar, A

    Jay, N., Rotman, N. H., Godfrey, P., Schapira, M., and Tamar, A. Internet congestion control via deep reinforcement learning. arXiv preprint arXiv:1810.03259 (2018)

    2018 arXiv

  35. [43]

    A., Jeong, H., Ihm, S., Han, D., and Park, K

    Jeong, E., Woo, S., Jamshed, M. A., Jeong, H., Ihm, S., Han, D., and Park, K. mtcp: a highly scalable user-level tcp stack for multicore systems. In NSDI (2014), pp. 489–502

    2014

  36. [44]

    Cfa: A practical prediction system for video qoe optimization

    Jiang, J., Sekar, V., Milner, H., Shepherd, D., Stoica, I., and Zhang, H. Cfa: A practical prediction system for video qoe optimization. In NSDI (2016), pp. 137–150

    2016

  37. [45]

    Pytheas: Enabling data-driven quality of experience optimization using group-based exploration-exploitation

    Jiang, J., Sun, S., Sekar, V., and Zhang, H. Pytheas: Enabling data-driven quality of experience optimization using group-based exploration-exploitation. In 14th USENIX Symposium on Networked Systems Design and Implementation (NSDI 17) (Boston, MA, 2017), USENIX Association, p...

    2017

  38. [46]

    Tcp performance in mobile ad hoc networks connected to the internet.REVISTA CIENTÍFICA PERIÓDICA- TELECOMUNICAÇÕES 10, 1 (2007), 22–29

    Karlsson, J., and Kassler, A. Tcp performance in mobile ad hoc networks connected to the internet.REVISTA CIENTÍFICA PERIÓDICA- TELECOMUNICAÇÕES 10, 1 (2007), 22–29

    2007

  39. [47]

    Identifying and aggregating homogeneous ipv4 /24 blocks with hobbit

    Lee, Y., and Spring, N. Identifying and aggregating homogeneous ipv4 /24 blocks with hobbit. In Proceedings of the 2016 Internet Measurement Conference (New York, NY, USA, 2016), IMC ’16, ACM, pp. 151–165

    2016

  40. [48]

    Efficient tuning of online systems using Bayesian optimization

    Letham, B., Karrer, B., Ottoni, G., and Bakshy, E. Efficient tuning of online systems using Bayesian optimization

  41. [49]

    include/linux/hashtable.h

    Levin, S. include/linux/hashtable.h. https://github.com/torvalds/linux/ blob/master/include/linux/hashtable.h

  42. [50]

    R., and Meleis, W

    Li, W., Zhou, F., Chowdhury, K. R., and Meleis, W. M.Qtcp: Adaptive congestion control with reinforcement learning. IEEE Transactions on Network Science and Engineering (2018)

    2018

  43. [51]

    L., Liang, M

    Li, Z. L., Liang, M. C.-J., He, W., Zhu, L., Dai, W., Jiang, J., and Sun, G. Metis: Robustly tuning tail latencies of cloud systems. InATC (USENIX Annual Technical Conference) (July 2018), USENIX

    2018

  44. [52]

    H., Viswanathan, R., Calder, M., Akella, A., Mahajan, R., Padhye, J., and Zhang, M

    Liu, H. H., Viswanathan, R., Calder, M., Akella, A., Mahajan, R., Padhye, J., and Zhang, M. Efficiently delivering online services over integrated infrastructure. In NSDI (2016), vol. 1, p. 1

    2016

  45. [53]

    Balloon powered internet

    Loon, P. Balloon powered internet. https://x.company/loon/

  46. [54]

    Speedchecker

    Ltd., S. Speedchecker

  47. [55]

    A., and Bustamante, F

    Lu, D., Qiao, Y., Dinda, P. A., and Bustamante, F. E. Character- izing and predicting tcp throughput on the wide area network. In Distributed Computing Systems, 2005. ICDCS 2005. Proceedings. 25th IEEE International Conference on (2005), IEEE, pp. 414–424

    2005

  48. [56]

    Neural adaptive video streaming with pensieve

    Mao, H., Netravali, R., and Alizadeh, M. Neural adaptive video streaming with pensieve. In Proceedings of the Conference of the ACM Special Interest Group on Data Communication (2017), ACM, pp. 197– 210

    2017

  49. [57]

    D., Beckman, R

    McKay, M. D., Beckman, R. J., and Conover, W. J. Comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics 21, 2 (1979), 239–245

    1979

  50. [58]

    Mike Belshe, R. P. Spdy protocol. https://tools.ietf.org/id/ draft-mbelshe-httpbis-spdy-00.txt

  51. [59]

    Network Information API, 2018

    Mozilla. Network Information API, 2018

    2018

  52. [60]

    K., Naylor, D., Jiang, J., Han, D., Seshan, S., and Zhang, H

    Mukerjee, M. K., Naylor, D., Jiang, J., Han, D., Seshan, S., and Zhang, H. Practical, real-time centralized control for cdn-based live video delivery. ACM SIGCOMM Computer Communication Review 45 , 4 (2015), 311–324

    2015

  53. [61]

    Inspectorgadget: Inferring network pro- tocol configuration for web services

    Naseer, U., and Benson, T. Inspectorgadget: Inferring network pro- tocol configuration for web services. In 2018 IEEE 38th International Conference on Distributed Computing Systems (ICDCS) (July 2018), pp. 1624–1629

    2018

  54. [62]

    An in-depth study of mobile browser performance

    Nejati, J., and Balasubramanian, A. An in-depth study of mobile browser performance. In Proceedings of the 25th International Con- ference on World Wide Web (2016), International World Wide Web Conferences Steering Committee, pp. 1305–1315

    2016

  55. [63]

    Polaris: Faster page loads using fine-grained dependency tracking

    Netravali, R., Goyal, A., Mickens, J., and Balakrishnan, H. Polaris: Faster page loads using fine-grained dependency tracking. In 13th USENIX Symposium on Networked Systems Design and Implementation (NSDI 16) (2016), USENIX Association

    2016

  56. [64]

    Tcp wise: One initial congestion window is not enough

    Nie, X., Zhao, Y., Chen, G., Sui, K., Chen, Y., Pei, D., Zhang, M., and Zhang, J. Tcp wise: One initial congestion window is not enough. In Performance Computing and Communications Conference (IPCCC), 2017 IEEE 36th International (2017), IEEE, pp. 1–8

    2017

  57. [65]

    Reducing 14 ConfigTron: Tackling network diversity with heterogeneous configurations , web latency through dynamically setting tcp initial window with reinforcement learning

    Nie, X., Zhao, Y., Pei, D., Chen, G., Sui, K., and Zhang, J. Reducing 14 ConfigTron: Tackling network diversity with heterogeneous configurations , web latency through dynamically setting tcp initial window with reinforcement learning

  58. [66]

    of Sheffield, M. L. G. U. GPyOpt

  59. [67]

    Bing and google agree: Slow pages lose users

    oreilly.com. Bing and google agree: Slow pages lose users. http: //radar.oreilly.com/2009/06/bing-and-google-agree-slow-pag.html

    2009

  60. [68]

    E., and Cantú-Paz, E

    Pelikan, M., Goldberg, D. E., and Cantú-Paz, E. Boa: The bayesian optimization algorithm. In Proceedings of the 1st Annual Conference on Genetic and Evolutionary Computation-Volume 1 (1999), Morgan Kaufmann Publishers Inc., pp. 525–532

    1999

  61. [69]

    Ap-atoms: A high-accuracy data-driven client aggregation for global load balancing

    Pi, Y., Jamin, S., Danzig, P., and Shaha, J. Ap-atoms: A high-accuracy data-driven client aggregation for global load balancing. IEEE/ACM Transactions on Networking 26, 6 (2018), 2748–2761

    2018

  62. [70]

    picoTCP. picotcp. http://www.picotcp.com/

  63. [71]

    Ruamviboonsuk, V., Netravali, R., Uluyol, M., and Madhyastha, H. V. Vroom: Accelerating the mobile web with server-aided depen- dency resolution. In Proceedings of the Conference of the ACM Special Interest Group on Data Communication (2017), ACM, pp. 390–403

    2017

  64. [72]

    Iroko: A framework to prototype reinforcement learning for data center traffic control

    Ruffy, F., Przystupa, M., and Beschastnikh, I. Iroko: A framework to prototype reinforcement learning for data center traffic control. arXiv preprint arXiv:1812.09975 (2018)

    2018 arXiv

  65. [73]

    Large-scale scanning of tcp’s initial window

    Rüth, J., Bormann, C., and Hohlfeld, O. Large-scale scanning of tcp’s initial window. In Proceedings of the 2017 Internet Measurement Conference (2017), ACM, pp. 304–310

    2017

  66. [74]

    An empirical view on content provider fairness

    Rüth, J., Kunze, I., and Hohlfeld, O. An empirical view on content provider fairness. arXiv preprint arXiv:1905.07152 (2019)

    2019 arXiv

  67. [75]

    Congestion-control throwdown

    Schapira, M., and Winstein, K. Congestion-control throwdown. In Proceedings of the 16th ACM Workshop on Hot Topics in Networks (2017), ACM, pp. 122–128

    2017

  68. [76]

    Decision Trees

    scikit learn.org. Decision Trees

  69. [77]

    V., Krishnamurthy, S

    Singh, S., Madhyastha, H. V., Krishnamurthy, S. V., and Govindan, R. Flexiweb: Network-aware compaction for accelerating mobile web transfers. In Proceedings of the 21st Annual International Conference on Mobile Computing and Networking (2015), ACM, pp. 604–616

    2015

  70. [78]

    Large sample properties of simulations using latin hypercube sampling

    Stein, M. Large sample properties of simulations using latin hypercube sampling. Technometrics 29, 2 (1987), 143–151

    1987

  71. [79]

    Raptor: Routing attacks on privacy in tor

    Sun, Y., Edmundson, A., V anbever, L., Li, O., Rexford, J., Chiang, M., and Mittal, P. Raptor: Routing attacks on privacy in tor

  72. [80]

    Cs2p: Improving video bitrate selection and adaptation with data-driven throughput prediction

    Sun, Y., Yin, X., Jiang, J., Sekar, V., Lin, F., W ang, N., Liu, T., and Sinopoli, B. Cs2p: Improving video bitrate selection and adaptation with data-driven throughput prediction. In Proceedings of the 2016 ACM SIGCOMM Conference (2016), ACM, pp. 272–285

    2016

  73. [81]

    Home network or access link? locating last-mile downstream throughput bottlenecks

    Sundaresan, S., Feamster, N., and Teixeira, R. Home network or access link? locating last-mile downstream throughput bottlenecks. In International Conference on Passive and Active Network Measurement (2016), Springer, pp. 111–123

    2016

  74. [82]

    Beyond socket options: making the linux tcp stack truly extensible

    Tran, V.-H., and Bonaventure, O. Beyond socket options: making the linux tcp stack truly extensible. arXiv preprint arXiv:1901.01863 (2019)

    2019 arXiv

  75. [83]

    A., and Uhlig, S

    Turkovic, B., Kuipers, F. A., and Uhlig, S. Interactions between congestion control algorithms. network 3, 17

  76. [84]

    On the stationarity of tcp bulk data transfers

    Urvoy-Keller, G. On the stationarity of tcp bulk data transfers. In International Workshop on Passive and Active Network Measurement (2005), Springer, pp. 27–40

    2005

  77. [85]

    J., and Zhang, B

    V an Aken, D., Pavlo, A., Gordon, G. J., and Zhang, B. Automatic database management system tuning through large-scale machine learning. In Proceedings of the 2017 ACM International Conference on Management of Data (2017), ACM, pp. 1009–1024

    2017

  78. [86]

    S., Balasubramanian, A., Krishnamurthy, A., and Wetherall, D

    W ang, X. S., Balasubramanian, A., Krishnamurthy, A., and Wetherall, D. Demystifying page load performance with wprof. In NSDI (2013), pp. 473–485

    2013

  79. [87]

    S., Balasubramanian, A., Krishnamurthy, A., and Wetherall, D

    W ang, X. S., Balasubramanian, A., Krishnamurthy, A., and Wetherall, D. How speedy is spdy? In NSDI (2014), pp. 387–399

    2014

  80. [88]

    S., Krishnamurthy, A., and Wetherall, D

    W ang, X. S., Krishnamurthy, A., and Wetherall, D. Speeding up web page loads with shandian. In NSDI (2016), pp. 109–122

    2016

  81. [89]

    Tcp ex machina: computer- generated congestion control

    Winstein, K., and Balakrishnan, H. Tcp ex machina: computer- generated congestion control. In ACM SIGCOMM Computer Commu- nication Review (2013), vol. 43, ACM, pp. 123–134

    2013

  82. [90]

    Y., Ma, J., Hill, G

    Y an, F. Y., Ma, J., Hill, G. D., Raghavan, D., W ahby, R. S., Levis, P., and Winstein, K. Pantheon: the training ground for internet congestion- control research. In 2018 USENIX Annual Technical Conference (USENIX ATC 18) (Boston, MA, 2018), USENIX Association, pp. 731–743

    2018

  83. [91]

    Tcp con- gestion avoidance algorithm identification

    Y ang, P., Shao, J., Luo, W., Xu, L., Deogun, J., and Lu, Y. Tcp con- gestion avoidance algorithm identification. IEEE/ACM Transactions on Networking (TON) 22, 4 (2014), 1311–1324

    2014

  84. [92]

    ZeroMQ. Zeromq. http://zeromq.org/

  85. [93]

    On the constancy of internet path properties

    Zhang, Y., and Duffield, N. On the constancy of internet path properties. In Proceedings of the 1st ACM SIGCOMM Workshop on Internet Measurement (2001), ACM, pp. 197–211

    2001

  86. [94]

    Bestconfig: tapping the performance potential of systems via automatic configuration tuning

    Zhu, Y., Liu, J., Guo, M., Bao, Y., Ma, W., Liu, Z., Song, K., and Y ang, Y. Bestconfig: tapping the performance potential of systems via automatic configuration tuning. In Proceedings of the 2017 Symposium on Cloud Computing (2017), ACM, pp. 338–350. 15

    2017

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.