Flows In Data Center Networks Research Articles

Intelligent load balancing in data center software‐defined networks

AbstractIn response to the increasing demand for efficient resource utilization in data center networks (DCNs), the development of intelligent load‐balancing algorithms has become crucial. This article introduces the dual double deep Q network (DDQN) algorithm, designed for software‐defined networking (SDN) environments within data centers. By leveraging deep reinforcement learning, DDQN addresses the challenges posed by dynamic traffic patterns, diverse flow requirements, and the coexistence of elephant and mice flows. Our algorithm adopts a comprehensive SDN approach, evaluating the network's status by analyzing switch load and bandwidth utilization. Using convolutional neural networks for elephant and mice flows in DCN, our algorithm enables adaptive learning and training tailored to the specific demands of elephant flows. Employing a double deep Q network architecture (DDQN), DDQN optimizes paths for both elephant and mice flows independently. Real‐time adaptation mechanisms make routing decisions based on the robust learning capabilities of DDQN, enhancing network utilization and reducing packet loss by generating optimal forwarding paths according to the current network state and traffic patterns. Simulations conducted in a Mininet environment with RYU as the controller, utilizing a fat‐tree data center topology, validate the efficacy of DDQN. The results demonstrate its effectiveness in achieving higher throughput, lower latency, and superior load balancing compared to traditional algorithms like equal‐cost multipath and Hedera.

HPSTOS: High-Performance and Scalable Traffic Optimization Strategy for Mixed Flows in Data Center Networks

In data center networks, traffic needs to be distributed among different paths using traffic optimization strategies for mixed flows. Most of the existing strategies consider either distributed or centralized mechanisms to optimize the latency of mice flows or the throughput of elephant flows. However, low network performance and scalability issues are intrinsic limitations of both strategies. In addition, the current elephant flow detection methods are inefficient. In this article, we propose a high-performance and scalable traffic optimization strategy (HPSTOS) based on a hybrid approach that leverages the advantages of both centralized and distributed mechanisms. HPSTOS improves the efficiency of elephant flow detection through sampling and flow-table identification. HPSTOS guarantees preferential transmission of mice flows using priority scheduling and adjusts their transmission rate by coding-based congestion control on the end-host, reducing their latency. Additionally, HPSTOS schedules elephant flows by cost-aware dynamic flow scheduling on a centralized controller to improve their throughput. The controller handles only elephant flows, which constitutes the minority of the flows, allowing effective scalability. Evaluations show that HPSTOS outperforms existing schemes by realizing efficient elephant flow detection and improving network performance and scalability.

An Approach Based on Knowledge-Defined Networking for Identifying Heavy-Hitter Flows in Data Center Networks

Heavy-Hitters (HHs) are large-volume flows that consume considerably more network resources than other flows combined. In SDN-based DCNs (SDDCNs), HHs cause non-trivial delays for small-volume flows known as non-HHs that are delay-sensitive. Uncontrolled forwarding of HHs leads to network congestion and overall network performance degradation. A pivotal task for controlling HHs is their identification. The existing methods to identify HHs are threshold-based. However, such methods lack a smart system that efficiently identifies HH according to the network behaviour. In this paper, we introduce a novel approach to overcome this lack and investigate the feasibility of using Knowledge-Defined Networking (KDN) in HH identification. KDN by using Machine Learning (ML), allows integrating behavioural models to detect patterns, like HHs, in SDN traffic. Our KDN-based approach includes mainly three modules: HH Data Acquisition Module (HH-DAM), Data ANalyser Module (HH-DANM), and APplication Module (HH-APM). In HH-DAM, we present the flowRecorder tool for organizing packets into flows records. In HH-DANM, we perform a cluster-based analysis to determine an optimal threshold for separating HHs and non-HHs. Finally, in HH-APM, we propose the use of MiceDCER for routing non-HHs efficiently. The per-module evaluation results corroborate the usefulness and feasibility of our approach for identifying HHs.

Performance evaluation of elephant flow predictors in data center networking

The dynamic characteristics and unpredictability of the traffic generated by applications and services have given rise to some problems in data center networks (DCN). In particular, the concentration of elephant flows in DCNs can lead to a load imbalance because the equal-cost multi-path (ECMP) routing maps them to the same path. This can result in packet discard, which generates packet re-transmission, causes additional latency, and further degrades link performance. In this paper, we evaluate the performance of different elephant flow predictors with the use of real DCN traces. A traffic time series from a Facebook data center is used to model the elephant flow predictors on a short-term basis and to provide the descriptive statistics and inferences the flows. The paper proposes a hybrid prediction model by combining aspects of the FARIMA and the Recurrent Neural Network (FARIMA-RNN) models. A methodology based on the rank of the prediction accuracy metrics is applied to compare the performance of the hybrid model with the ARIMA, GARCH, RBF, MLP, and LSTM models. The comparison allows the selection of an appropriate model for the implementation of elephant flow scheduling techniques in DCNs. Results show that the FARIMA-RNN model presents lower error rates than the other predictors.

Buffer Occupancy-Based Transport to Reduce Flow Completion Time of Short Flows in Data Center Networks

Today’s data centers host a variety of different applications that impose specific requirements for their flows. Applications that generate short flows are usually latency sensitive; they require their flows to be completed as fast as possible. Short flows suffer to quickly increase their sending rate due to the existing long flows occupying most of the available capacity. This problem is caused due to the slow convergence of the current data center transport protocols. In this paper, we present a buffer occupancy-based transport protocol (BOTCP) to reduce flow completion time of short flows. BOTCP consists of two parts: (i) A novel buffer occupancy-based congestion signal, and (ii) a congestion control scheme that uses the congestion signal to reduce flow completion time of short flows. The proposed buffer occupancy-based congestion signal gives a precise measure of congestion. The congestion control scheme makes a differentiated treatment of short and long flows to reduce flow completion time of short flows. The results show that BOTCP significantly improves flow completion time of short flows over the existing transport protocols in data center networks.

Methods for Predicting Behavior of Elephant Flows in Data Center Networks

Several Traffic Engineering (TE) techniques based on SDN (Software-defined networking) proposed to resolve flow competitions for network resources. However, there is no comprehensive study on the probability distribution of their throughput. Moreover, there is no study on predicting the future of elephant flows. To address these issues, we propose a new stochastic performance evaluation model to estimate the loss rate of two state-of-art flow scheduling algorithms including Equalcost multi-path routing (ECMP), Hedera besides a flow congestion control algorithm which is Data Center TCP (DCTCP). Although these algorithms have theoretical and practical benefits, their effectiveness has not been statistically investigated and analyzed in conserving the elephant flows. Therefore, we conducted extensive experiments on the fat-tree data center network to examine the efficiency of the algorithms under different network circumstances based on Monte Carlo risk analysis. The results show that Hedera is still risky to be used to handle the elephant flows due to its unstable throughput achieved under stochastic network congestion. On the other hand, DCTCP found suffering under high load scenarios. These outcomes might apply to all data center applications, in particular, the applications that demand high stability and productivity.

A Dynamical and Load-Balanced Flow Scheduling Approach for Big Data Centers in Clouds

Load-balanced flow scheduling for big data centers in clouds, in which a large amount of data needs to be transferred frequently among thousands of interconnected servers, is a key and challenging issue. The OpenFlow is a promising solution to balance data flows in data center networks through its programmatic traffic controller. Existing OpenFlow based scheduling schemes, however, statically set up routes only at the initialization stage of data transmissions, which suffers from dynamical flow distribution and changing network states in data centers and often results in poor system performance. In this paper, we propose a novel dynamical load-balanced scheduling (DLBS) approach for maximizing the network throughput while balancing workload dynamically. We first formulate the DLBS problem, and then develop a set of efficient heuristic scheduling algorithms for the two typical OpenFlow network models, which balance data flows time slot by time slot. Experimental results demonstrate that our DLBS approach significantly outperforms other representative load-balanced scheduling algorithms Round Robin and LOBUS; and the higher imbalance degree data flows in data centers exhibit, the more improvement our DLBS approach will bring to the data centers.

A New Approach to Stochastic Scheduling in Data Center Networks

The Quality of Service (QoS) of scheduling between latency-sensitive small data flows (a.k.a. mice) and throughput-oriented large ones (a.k.a. elephants) has become ever challenging with the proliferation of cloud-based applications. In light of this mounting problem, this work proposes a novel flow control scheme, HOLMES (HOListic Mice-Elephants Stochastic), which offers a holistic view of global congestion awareness as well as a stochastic scheduler of mixed mice-elephants data flows in Data Center Networks (DCNs). Firstly, we theoretically prove the necessity for partitioning DCN paths into sub-networks using a stochastic model. Secondly, the HOLMES architecture is proposed, which adaptively partitions the available DCN paths into low-latency and high-throughput sub-networks via a global congestion-aware scheduling mechanism. Based on the stochastic power-of-two-choices policy, the HOLMES scheduling mechanism acquires only a subset of the global congestion information, while achieves close to optimal load balance on each end-to-end DCN path. We also formally prove the stability of HOLMES flow scheduling algorithm. Thirdly, extensive simulation validates the effectiveness and dependability of HOLMES with select DCN topologies. The proposal has been in test in an industrial production environment. An extensive survey of related work is also presented.

Fast Coupled Retransmission for Multipath TCP in Data Center Networks

Multipath TCP (MPTCP) is emerging as a new potential candidate that replaces the existing single-path TCP (SPTCP). One applicable area of MPTCP is cloud data center networks, where multiple paths are often available between servers. However, it is observed that MPTCP may not fully leverage the multiple paths and sometimes perform even worse than SPTCP, especially for time-sensitive short flows in data center networks. To address this challenge, we propose a fast coupled retransmission mechanism for MPTCP. The key observation is that one of the paths is likely to be congested if an out-of-order packet exists even after all data packets have been transmitted. In this case, we forward the data packets from the congested path onto the noncongested path and quickly retransmit them. This mechanism effectively avoids network congestion and, hence, helps reduce the overall flow completion time. We implement the proposed scheme in the Linux kernel and evaluate it on our real data center testbed, and the results show that our proposal significantly outperforms the existing schemes.

ATME: Accurate Traffic Matrix Estimation in Both Public and Private Datacenter Networks

Understanding the pattern of end-to-end traffic flows in datacenter networks (DCNs) is essential to many DCN designs and operations (e.g., traffic engineering and load balancing). However, little research work has been done to obtain traffic information efficiently and yet accurately. Researchers often assume the availability of traffic tracing tools (e.g., OpenFlow) when their proposals require traffic information as input, but these tools may have high monitoring overhead and consume significant switch resources even if they are available in a DCN. Although estimating the traffic matrix (TM) between origin-destination pairs using only basic switch SNMP counters is a mature practice in IP networks, traffic flows in DCNs show totally different characteristics, while the large number of redundant routes in a DCN further complicates the situation. To this end, we propose to utilize resource provisioning information in public cloud datacenters and the service placement information in private datacenters for deducing the correlations among top-of-rack switches, and to leverage the uneven traffic distribution in DCNs for reducing the number of routes potentially used by a flow. These allow us to develop ATME as an efficient TM estimation scheme that achieves high accuracy for both public and private DCNs. We compare our two algorithms with two existing representative methods through both experiments and simulations; the results strongly confirm the promising performance of our algorithms.

SDN에서 엘리펀트 플로우의 실시간 분류, 시각화 및 QoS 제어

오래 지속되는 플로우인 엘리펀트 플로우는 데이터 센터에서 많은 대역폭을 소비해서, 지연에 민감하고 짧은 시간 흐르는 플로우인 마이스 플로우의 흐름을 방해하게 된다. 이는 마이스 플로우에 대해 적지않은 지연을 발생시켜서, 결과적으로 그 네트워크에서 동작하고 있는 응용의 성능을 저하시키게 된다. 그러므로 데이터 센터 네트워크는 이를 분류하고 시각화 해내어 실시간으로 QoS 프로비져닝을 제공할 수 있어야 한다. 본 논문에서는 다음에 대해 논한다. (1) SDN에서 sflow를 이용한 엘리펀트 플로우의 실시간 검출 및 시각화를 위한 프레임워크 제시. 이는 시각화된 토폴로지에서 스위치를 더블클릭 하므로써 스위치를 통과하는 엘리펀트 플로우를 점검할 수 있게 한다. (2) SDN 제어기에 의해서 정의되고 관리되는 QoS를 보장하는 접근 및 OpenFlow에 의해 제공되는 규격. 본 논문에서는 SDN 네트워크내에서 rate-limiting (traffic-shaping) 분류 기법을 사용하는 것을 주로 논의한다. Long-lived flowed termed as elephant flows in data center networks have a tendency to consume a lot of bandwidth, leaving delay-sensitive short-lived flows referred to as mice flows choked behind them. This results in non-trivial delays for mice flows, eventually degrading application performance running on the network. Therefore, a datacenter network should be able to classify, detect, and visualize elephant flows as well as provide QoS guarantees in real-time. In this paper we aim to focus on: 1) a proposed framework for real-time detection and visualization of elephant flows in SDN using sFlow. This allows to examine elephant flows traversing a switch by double-clicking the switch node in the topology visualization UI; 2) an approach to guarantee QoS that is defined and administered by a SDN controller and specifications offered by OpenFlow. In the scope of this paper, we will focus on the use of rate-limiting (traffic-shaping) classification technique within an SDN network.

FSQCN: Fast and simple quantized congestion notification in data center ethernet

Currently, Quantized Congestion Notification (QCN) has been accepted as the standard layer 2 congestion control protocol in Data Center Ethernet. In this study, we find that QCN has two drawbacks. First, the incomplete binary search of QCN may fail to find the proper sending rate, leading to the complicate supplement mechanisms of QCN. Accordingly, the cost of hardware implementation is increased. Second, in face of unknown network environments, the rate setting of QCN is inconsistent. More specifically, the sending rate is set to be the link rate at initialization, while QCN additively probes for extra bandwidth in face of dynamic network. Consequently, QCN may spend much time on acquiring the spare bandwidth if its sending rate is increased from a low value. To address these problems, we propose the Fast and Simple QCN (FSQCN), following the same framework as QCN. FSQCN complements the binary search and removes other complicate supplement mechanisms in QCN. Thus, FSQCN is much simpler and more efficient than QCN. Moreover, FSQCN resets the sending rate to the link rate explicitly when the switch detects spare bandwidth. Extensive simulation experiments validate that FSQCN controls the queue length well like QCN, and responds faster than QCN, especially in face of the large number of burst short flows in data center networks.

Make‐way: transporting latency‐sensitive flows nonblockingly in oversubscription data center networks

SummaryMany recent studies show that data center networks usually mix with a large amount of latency‐agnostic background flows and a large number of latency‐sensitive flows. The former, although contributing most of the traffic, is deadline‐agnostic, while the latter always confronts a soft‐real‐time constraint, namely deadline. Missing this deadline may cause performance impairment. For this reason, directly using the traditional TCP in data center networks, which is deadline‐agnostic, may suffer from performance and efficiency problems. Recently, many works improve on TCP to address this problem. But all of them focus on the latency‐sensitive flows themselves. Because latency‐sensitive flows, although large in number, only account for a small part of traffic, these methods cannot effectively ensure deadline for the latency‐sensitive flows. In this paper, we propose Make‐way, a new data center network transport protocol for satisfying the deadlines of latency‐sensitive flows. In Make‐way, just as what the name depicts, once a latency‐sensitive flow encounters congestion, the latency‐agnostic background flows will make way for it. Especially, Make‐way does not need any special hardware support of modification. For the latency‐agnostic flows in data center networks usually contribute the majority traffic, by doing so, the latency‐sensitive flows may be transported nonblockingly in data center networks and thus can meet their deadlines. Extensive simulation results show that Make‐way can meet the deadlines of latency‐sensitive flows with a probability of more than 97%. Copyright © 2015 John Wiley & Sons, Ltd.

Schemes for Fast Transmission of Flows in Data Center Networks

In this paper, we survey different existing schemes for the transmission of flows in Data Center Networks (DCNs). The transport of flows in DCNs must cope with the bandwidth demands of the traffic that a large number of data center applications generates and achieve high utilization of the data center infrastructure to make the data center financially viable. Traffic in DCNs roughly comprises short flows, which are generated by the Partition/Aggregate model adopted by several applications and have sizes of a few kilobytes, and long flows, which are data for the operation and maintenance of the data center and have sizes on the order of megabytes. Short flows must be transmitted (or completed) as soon as possible or within a deadline, and long flows must be serviced with a minimum acceptable throughput. The coexistence of short and long flows may jeopardize achieving both performance objectives simultaneously. This challenge has motivated growing research on schemes for managing the transmission of flows in DCNs. We describe several recent schemes aimed at reducing the flow completion time in DCNs. We also present a summary of existing solutions for the incast traffic phenomenon. We provide a comparison and classification of the surveyed schemes, describe their advantages and disadvantages, and show the different trends for scheme design. For completeness, we describe some DCN architectures, discuss the traffic patterns of DCNs, and discuss why some existing versions of transport protocols may not be usable in DCNs. At the end, we discuss some of the identified research challenges.

Coarse-grained traffic matrix estimation for data center networks

We address the problem of estimating the real-time traffic flows in data center networks (DCNs), using the light-weight SNMP data. Unlike the problem of estimating the traffic matrix (TM) across origin–destination (OD) pairs in ISP networks, the traffic flows across servers or Top of Rack (ToR) switch pairs in DCNs are notoriously more irregular and volatile. Although numerous methods have been proposed in past several years to solve the TM estimation problem in ISP networks, none of them could be applied to DCNs directly. In this paper, we make the first step to solve the TM estimation problem in DCNs by leveraging the characteristics of prevailing data center architectures and decomposing the topologies of DCNs, which makes TM estimation problems in DCNs easy to handle. We also state a basic theory to obtain the aggregate traffic characteristics of these clusters unbiasedly. We propose two efficient TM estimation algorithms based on the decomposed topology and the aggregate traffic information, which improves the state-of-the-art tomography methods without requiring any additional instrumentation. Finally, we compare our proposal with a recent representative TM estimation algorithm through both real experiments and extensive simulations, the results show that, (i) the data center TM estimation problem could be well handled after the decomposition step, (ii) our two algorithms outperforms the former one in both speed and accuracy.