Timely

Mittal, Radhika; Lam, Vinh The; Dukkipati, Nandita; Blem, Emily; Wassel, Hassan M. G.; Ghobadi, Monia; Vahdat, Amin; Wang, Yaogong; Wetherall, David; Zats, David

doi:10.1145/2829988.2787510

Cited by 213 publications

(29 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Advanced CC mechanisms for DCNs: Some recent pioneering works have proposed sophisticated CC mechanisms showing outstanding performance in DCNs [3,23,25,27,29,47]. For example, HPCC [25] achieves remarkably short FCTs while offering good throughput and stability to traffic changes (e.g., incast events).…”

Section: Related Workmentioning

confidence: 99%

“…The last decade has witnessed an ever-growing interest on optimizing Data Center Networks (DCN) given the critical services they run and the high CAPEX and OPEX these infrastructures entail. Among the large spectrum of traffic optimization mechanisms [24], Congestion Control (CC) has been especially explored in the literature [3,23,25,27,29,44,47]. Nowadays, most production DCNs run two main CC protocols: (𝑖) DCTCP [3], for datacenters based on the traditional TCP/IP stack, and (𝑖𝑖) DCQCN [47], for emerging RDMA-based DCNs.…”

Section: Introductionmentioning

confidence: 99%

“…Congestion Control (CC) has been extensively studied in the past. As a result, there exists a plethora of pioneering solutions for DCNs tackling the problem from different angles, such as RTTbased [23,27], credit-based [12,29], or telemetry-based [7,25] mechanisms. Nowadays most production DCNs implement two main well-established CC protocols: DCTCP [3], and DCQCN [47].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

MAGNNETO: A Graph Neural Network-Based Multi-Agent System for Traffic Engineering

Bernárdez¹,

Suárez-Varela²,

López³

et al. 2023

IEEE Trans. Cogn. Commun. Netw.

View full text Add to dashboard Cite

Current trends in networking propose the use of Machine Learning (ML) for a wide variety of network optimization tasks. As such, many efforts have been made to produce ML-based solutions for Traffic Engineering (TE), which is a fundamental problem in ISP networks. Nowadays, state-of-the-art TE optimizers rely on traditional optimization techniques, such as Local search, Constraint Programming, or Linear programming. In this paper, we present MAGNNETO, a distributed ML-based framework that leverages Multi-Agent Reinforcement Learning and Graph Neural Networks for distributed TE optimization. MAGNNETO deploys a set of agents across the network that learn and communicate in a distributed fashion via message exchanges between neighboring agents. Particularly, we apply this framework to optimize link weights in OSPF, with the goal of minimizing network congestion. In our evaluation, we compare MAGNNETO against several state-of-the-art TE optimizers in more than 75 topologies (up to 153 nodes and 354 links), including realistic traffic loads. Our experimental results show that, thanks to its distributed nature, MAGNNETO achieves comparable performance to state-of-the-art TE optimizers with significantly lower execution times. Moreover, our ML-based solution demonstrates a strong generalization capability to successfully operate in new networks unseen during training.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

MAGNNETO: A Graph Neural Network-Based Multi-Agent System for Traffic Engineering

Bernárdez¹,

Suárez-Varela²,

López³

et al. 2023

IEEE Trans. Cogn. Commun. Netw.

View full text Add to dashboard Cite

show abstract

“…For example, multiple iterations of CC have been deployed in the wild, since TCP Tahoe, including TCP Cubic [18], TCP Compound [51] and MultiPath TCP [43]. At the same time, there has been a plethora of proposed CC algorithms for data centre ( [5,36,61]), wireless ([26, 35, 47]) and satellite ([3, 14, 54]) networks all of which present wildly different characteristics and challenges compared to each other.…”

Section: Congestion Controlmentioning

confidence: 99%

RayNet: A Simulation Platform for Developing Reinforcement Learning-Driven Network Protocols

Giacomoni¹,

Benny²,

Parisis³

2023

Preprint

View full text Add to dashboard Cite

Reinforcement Learning has gained significant momentum in the development of network protocols. However, learning-based protocols are still in their infancy, and substantial research is required to build deployable solutions. Developing a protocol based on reinforcement learning is a complex and challenging process that involves several model design decisions and requires significant training and evaluation in real or realistic network topologies. Network simulators offer RL-based protocols a highly effective training environment, because simulations are deterministic and can run in parallel. In this paper, we introduce RayNet, a scalable and adaptable simulation framework for the development of learning-based network protocols. RayNet integrates OMNeT++, a fully programmable network simulator, with Ray/RLlib, a scalable training platform for distributed reinforcement learning. RayNet facilitates the methodical development of RL-based network protocols with minimal overhead. We have developed a congestion control use case and present evidence that RayNet can be a valuable framework for the computer networks research community.

show abstract

“…According to RFC 793 [7], Tahoe [8], Reno [9], and NewReno [10], binary increase congestion control (BIC) [11] and Cubic [12] are considered loss-based TCP-CCAs. Vegas [13], FAST [14], low latency congestion control (Lola) [15], and Timely [16] are known as delay-based TCP-CCAs. Vegas-reno (Veno) [17], Compound [18], Fusion [19], Illinois [20], bottleneck bandwidth and round trip time (BBR) [21], and performance-oriented congestion control (PCC) [22] are considered hybrid TCP-CCAs.…”

mentioning

confidence: 99%

D-OLIA: A Hybrid MPTCP Congestion Control Algorithm with Network Delay Estimation

Lubna

Mahmud

Kim

et al. 2021

Sensors

View full text Add to dashboard Cite

With the recent evolution of mobile technology, modern devices equipped with multiple communication interfaces have become popular. The multipath transmission control protocol (MPTCP) has evolved to facilitate multiple communication interfaces through a single TCP connection for faster Internet access. MPTCP congestion control algorithms (MPTCP-CCAs) control data flow by fulfilling three design goals, i.e., ensuring improvement over single-path flows, ensuring fairness, and balancing congestion. Current MPTCP-CCAs cannot fulfill these design goals. For example, the opportunistic-linked increase algorithm (OLIA), a well-known MPTCP-CCA in load balancing, often results in low throughput because it cannot properly utilize the underlying network. In addition, the current Internet has a rapidly changing characteristic due to a large amount of short-lived traffic, making it difficult for MPTCP-CCAs to cope. An awareness of prevailing network delay conditions might help MPTCP-CCAs to utilize the network capacity fully. Therefore, we propose dynamic OLIA (D-OLIA), a hybrid MPTCP-CCA that enhances the performance of OLIA by integrating an awareness of the current network delay condition for deciding the congestion window (CWND) decrease factor. We estimate the current network delay condition, i.e., less-congested or congested, by observing the changes in the round-trip-time (RTT). Based on the estimated network delay condition, we decide the CWND decrease factor in real-time for reducing the CWND during packet loss events. We implemented D-OLIA in the Linux kernel and experimented using the Mininet emulator. The emulation results demonstrate that D-OLIA successfully estimates current network delay conditions and results in approximately a 20% increased throughput compared to the original OLIA. Compared to certain MPTCP-CCAs, it also yields a highly improved performance in terms of throughput, RTT, packet retransmissions, and fairness among the MPTCP sub-flows.

show abstract

Timely

Cited by 213 publications

References 37 publications

MAGNNETO: A Graph Neural Network-Based Multi-Agent System for Traffic Engineering

MAGNNETO: A Graph Neural Network-Based Multi-Agent System for Traffic Engineering

RayNet: A Simulation Platform for Developing Reinforcement Learning-Driven Network Protocols

D-OLIA: A Hybrid MPTCP Congestion Control Algorithm with Network Delay Estimation

Contact Info

Product

Resources

About