Deficit Round Robin Research Articles

Deficit Round-Robin: A Second Network Calculus Analysis

Deficit Round-Robin (DRR) is a widespread scheduling algorithm that provides fair queueing with variable-length packets. Bounds on worst-case delays for DRR were found by Boyer et al., who used a rigorous network calculus approach and characterized the service obtained by one flow of interest by means of a convex strict service curve. These bounds do not make any assumptions on the interfering traffic hence are pessimistic when the interfering traffic is constrained by some arrival curves. For such cases, two improvements were proposed. The former, by Soni et al., uses a correction term derived from a semi-rigorous heuristic; unfortunately, these bounds are incorrect, as we show by exhibiting a counter-example. The latter, by Bouillard, rigorously derive convex strict service curves for DRR that account for the arrival curve constraints of the interfering traffic. In this paper, we improve on these results in two ways. First, we derive a non-convex strict service curve for DRR that improves on Boyer et al. when there is no arrival constraint on the interfering traffic. Second, we provide an iterative method to improve any strict service curve (including Bouillard’s) when there are arrival constraints for the interfering traffic. As of today, our results provide the best-known worst-case delay bounds for DRR. They are obtained by using the method of the pseudo-inverse.

Building a Digital Twin for network optimization using Graph Neural Networks

Network modeling is a critical component of Quality of Service (QoS) optimization. Current networks implement Service Level Agreements (SLA) by careful configuration of both routing and queue scheduling policies. However, existing modeling techniques are not able to produce accurate estimates of relevant SLA metrics, such as delay or jitter, in networks with complex QoS-aware queueing policies (e.g., strict priority, Weighted Fair Queueing, Deficit Round Robin). Recently, Graph Neural Networks (GNNs) have become a powerful tool to model networks since they are specifically designed to work with graph-structured data. In this paper, we propose a GNN-based network model able to understand the complex relationship between (i) the queueing policy (scheduling algorithm and queue sizes), (ii) the network topology, (iii) the routing configuration, and (iv) the input traffic matrix. We call our model TwinNet, a Digital Twin that can accurately estimate relevant SLA metrics for network optimization. TwinNet can generalize to its input parameters, operating successfully in topologies, routing, and queueing configurations never seen during training. We evaluate TwinNet over a wide variety of scenarios with synthetic traffic and validate it with real traffic traces. Our results show that TwinNet can provide accurate estimates of end-to-end path delays in 106 unseen real-world topologies, under different queuing configurations with a Mean Absolute Percentage Error (MAPE) of 3.8%, as well as a MAPE of 6.3% error when evaluated with a real testbed. We also showcase the potential of the proposed model for SLA-driven network optimization and what-if analysis.

Graph-neural-network-based delay estimation for communication networks with heterogeneous scheduling policies

Modeling communication networks to predict performance such as delay and jitter is important for evaluating and optimizing them. In recent years, neural networks have been used to do this, which may have advantages over existing models, for example from queueing theory. One of these neural networks is RouteNet, which is based on graph neural networks. However, it is based on simplified assumptions. One key simplification is the restriction to a single scheduling policy, which describes how packets of different flows are prioritized for transmission. In this paper we propose a solution that supports multiple scheduling policies (Strict Priority, Deficit Round Robin, Weighted Fair Queueing) and can handle mixed scheduling policies in a single communication network. Our solution is based on the RouteNet architecture as part of the "Graph Neural Network Challenge". We achieved a mean absolute percentage error under 1% with our extended model on the evaluation data set from the challenge. This takes neural-network-based delay estimation one step closer to practical use.

Individual Service Curves for Bandwidth-Sharing Policies Using Network Calculus

Bandwidth-sharing policies guarantee a proportion of the service offered by a server to each class of traffic. These policies include GPS (Generalized Processor Sharing) and DRR (Deficit Round Robin). In this letter, we compute service curves for each class of traffic for such scheduling policies using network calculus. These individual service curves take into account the characteristics of the cross-traffic and thus improve the state-of-the-art performance bounds like delay or backlog.

A Comparison of Fair Sharing Algorithms for Regulating Search as a Service API

Providers of a Search as a Service (SaaS) environment must ensure that their users will not monopolize the service or use more than their fair share of resources. Fair sharing algorithms have long been used in computer networking to balance access to a router or switch, and some of these algorithms have also been applied to the control of queries submitted to search engine APIs. If a search query’s execution cost can be reliably estimated, fair sharing algorithms can be applied to the input of a SaaS API to ensure everyone has equitable access to the search engine.
 The novelty of this paper lies in presenting a Single-Server Max-Min Fair Deficit Round Robin algorithm, a modified version of the Multi-Server Max-Min Fair Deficit Round Robin algorithm. The Single-Server Max-Min Fair Deficit Round Robin algorithm is compared to three other fair sharing algorithms, token-bucket, Deficit Round Robin (DRR), and Peng and Plale’s [1] Modified Deficit Round Robin (MDRR) in terms of three different usage scenarios, balanced usage, unbalanced usage as well as an idle client usage, to determine which is the most suitable fair sharing algorithm for use in regulating traffic to a SaaS API. This research demonstrated that the Single-Server Max-Min Fair DRR algorithm provided the highest throughput of traffic to the search engine while also maintaining a fair balance of resources among clients by re-allocating unused throughput to clients with saturated queues so a max-min allocation was achieved.

A novel Equilibrated scheduling algorithm for multimedia transmission in Internet of Multimedia Things

Internet of Multimedia Things (IoMT) has recently witnessed a rapid growth in the world of new technologies that provide numerous multimedia-based services. The primary objective of this work is to develop a novel equilibrated extension of the Deficit Weighted Round Robin (DWRR) scheduling algorithm entitled EDWRR. The proposed algorithm incorporates a Medium Access Control (MAC) layer solution for the purpose of improving the Quality of Service (QoS) of multimedia transmission in IoMT networks. The results highlight the efficiency of the proposed algorithm with regards to the reduction of the packet drop rate. Moreover, the proposed algorithm provides an efficient queue management by taking advantage of the unused memory space. In addition, through the attempt to tackle the problem of overloaded queues, the proposed algorithm is able to achieve better performance in comparison with the other scheduling algorithms in terms of packet delivery ratio (PDR), average end-to-end delay and throughput.

A Smart Cloud Service Management Algorithm for Vehicular Clouds

Vehicular clouds (VCs) have become a promising research area due to its on-demand solutions, resource pooling, unified services, autonomous cloud formation and transformational management. It makes use of the underutilized resources of vehicles on the parking lot, roadways, driveways and streets, and creates the infrastructure to support various services offered by the cloud service provider (CSP) by deploying virtual machines (VMs). However, these vehicles can leave the coverage/grid of VC due to its mobility and change in the environment. Therefore, the hosted VMs on those vehicles can be transferred to other potential vehicles (i.e., migration) in order to avoid disruption of services. These services can be viewed as user requests (URs) submitted to the CSP by cloud users. Here, the challenging tasks are to map the URs to the VMs (or vehicles) and identify the potential vehicles for migration, and they need immediate attention. In this paper, we propose a smart cloud service management (SCSM) algorithm for VCs and address the above challenges. This algorithm consists of three phases, namely assignment of vehicles to grids, URs to grids and URs to vehicles by considering the mobility pattern of vehicles. The performance of SCSM is assessed using three traffic congestion scenarios and thirty-six instances of four datasets, and compared with round-robin (RR) and deficit weighted RR (DWRR) using seven performance metrics. The comparison results show that SCSM achieves 58% and 57% (33% and 33%) better than RR and DWRR in makespan (number of migrations) and other performance metrics.

Two-level scheduling scheme for integrated 4G-WLAN network

In this paper, a novel scheduling scheme for the Fourth Generation (4G)-Wireless Local Area Network (WLAN) network is proposed to ensure that end to end traffic transaction is provisioned seamlessly. The scheduling scheme is divided into two stages; in stage one, traffic is separated into Actual Time Traffic (ATT) and Non-Actual-Time Traffic (NATT), while in stage two, complex queuing strategy is performed. In stage one, Class-Based Queuing (CBQ) and Deficit Round Robin (DRR) are used for NATT and ATT applications, respectively to separate and forward traffic themselves according to source requirements. Whereas in the stage, two Control Priority Queuing (CPQ) is used to assign each class the appropriate priority level. Evaluation of the performance of the integrated network was done according to several metrics such as end-to-end delay, jitter, packet loss, and network’s throughput. Results demonstrate major improvements for AT services with minor degradation on NAT applications after implementing the new scheduling scheme.

Pengukuran Performa Algoritma Kendali Kongesti Active Queue Manajemen (AQM) Pada Jaringan Backbone Multicast Berbasis Protocol Multicast PIM-DM

UDP is a data transmission protocol on the Internet that allows data to be sent in realtime and become a development base of various services such as IPTV, VOIP, VOD, and video conferencing. The multicast network was a solution to provide better service to many public needs with the concept of simplicity and efficiency. Dense traffic situation triggering a packet loss and delay that not tolerated by some services, especially multimedia streaming service, while UDP does not guarantee the quality of service so that the necessary traffic management method that serves as a traffic controller to the congestion in the network. This study analyzes the performance of technology based on Active Queue Management (AQM) congestion control in a multicast backbone network based on Protocol Independent Multicast – Dense Multicast using Droptail, Deficit Round Robin (DRR), and Random Early Detection (RED) algorithm, which is simulated using NS2. The simulation results show that the congestion control algorithm, DRR provides the best performance with a balanced value of mean delay, mean throughput, and the mean loss. RED is a congestion control algorithm which has the best queue management mechanism but not able to suppress the mean loss in each scenario tested because RED has a small amount of buffer size, Droptail not an appropriate algorithm for multicast networks implemented in, the absence of an interrupt mechanism trigger data services will have a tendency towards a particular data packet.

Regulating Scheduler (RSC): A Novel Solution for IEEE 802.1 Time Sensitive Network (TSN)

Emerging applications such as industrial automation, in-vehicle, professional audio-video, and wide area electrical utility networks require strict bounds on the end-to-end network delay. Solutions so far to such a requirement are either impractical or ineffective. Flow based schedulers suggested in a traditional integrated services (IntServ) framework are O(N) or O(log N), where N is the number of flows in the scheduler, which can grow to tens of thousands in a core router. Due to such complexity, class-based schedulers are adopted in real deployments. The class-based systems, however, cannot provide bounded delays in networks with cycle, since the maximum burst grows infinitely along the cycled path. Attaching a regulator in front of a scheduler to limit the maximum burst is considered as a viable solution. International standards, such as IEEE 802.1 time sensitive network (TSN) and IETF deterministic network (DetNet) are adopting this approach as a standard. The regulator in TSN and DetNet, however, requires flow state information, therefore contradicts to the simple class-based schedulers. This paper suggests non-work conserving fair schedulers, called ‘regulating schedulers’ (RSC), which function as a regulator and a scheduler at the same time. A deficit round-robin (DRR) based RSC, called nw-DRR, is devised and proved to be both a fair scheduler and a regulator. Despite the lower complexity, the input port-based nw-DRR is shown to perform better than the current TSN approach, and to bind the end-to-end delay within a few milliseconds in realistic network scenarios.

Defense Against Protocol Level Attack in Tor Network using Deficit Round Robin Queuing Process

Among different types of network, Tor network is mostly referred as an onion-routing network which enables the anonymous communication and supports TCP applications over the network. The onion-routing is utilized by the Tor network for transmitting the information of users via virtual circuits which are created by several successive relays. The performance and security of the anonymous communication in Tor network have been improved by using Novel Traffic Dividing and Scheduling (NTDS) mechanism. Normally, this mechanism was used for preventing the basic versions of sniper attacks in the Tor network. On the other hand, protocol-level attacks were not possible to prevent since some malicious entry router may duplicate, alter, insert or delete the TCP stream cells from the sender. This may cause cell recognition errors at the exit router and these cell recognition errors were recognized at the destination which helps to identify the sender. Hence, in this article, a defense mechanism called Mid-DRRQ is proposed for protocol-level attacks against Tor network. In Mid-DRRQ mechanism, the protocol-level attacks are prevented by using indistinguishability obfuscation model which generates the fake traffic within the middle relay. This fake traffic is allowed on outgoing connections without any cover traffic. Moreover, the outgoing connections of middle relays are controlled by using Deficit Round Robin queuing process. Finally, the simulation results show that the proposed mechanism effectively prevents the protocol-level attacks against Tor network.

Quality of Service Impact on Deficit Round Robin and Stochastic Fair Queuing Mechanism in Wired-cum-Wireless Network

The deficient round robin (DRR) and stochastic fair queue (SFQ) are the active queue mechanism (AQM) techniques. These AQM techniques play important role in buffer management in order to control the congestion in the wired-cum-wireless network by dropping packets during the buffer overflow or near to overflow. This research study focus on the performance evaluation of the DRR and SFQ using different scenarios such as increasing number of node scenario, pause time scenario and mobility scenario. We evaluate the performance of DRR and SFQ based on two parameters such as average packet delay and average packet dropped. In case of increasing number of nodes, the SFQ has outperformed than DRR by having comparatively low per packet delay. DRR has higher packet dropped ratio as compare to SFQ. In mobility and pause time scenario, SFQ has less per packet delay while DRR has less packet dropped ratio These results revealed that DRR performance was affected by an increase in the number of nodes in a network. The DRR send the packet in a round-robin fashion without caring about the bandwidth of a path due to which the packet dropped ratio was high. On another hand, the SFQ has comparatively outperformed in all scenarios by having less per packet delay. SFQ become aggressive by dropping more data packets during buffer overflow. In short, SFQ will be preferred for a network where the congestion occurred more frequently.

Multipath Optimized Link State Routing Protocol with Multi-Variant Deficit Round Robin

Optimized Link State Routing Protocol (OLSR) is extended by multipath allocation of bandwidths which is known to be Multipath Optimized Link State Routing Protocol (MP-OLSR). OLSR is proactive that widely used in communication of dense and complex network traffic as it is more stable and easily adaptive. MP-OLSR takes multipath Dijkstra’s algorithm which is used for better performance of MANET routing protocols by considering scalability, route recovery and loop breakage detection in communication network traffic. MP-OLSR protocols with improvements have been proposed such as implementation of Round Robin, weighted Round Robin (WRR). But these protocols lack multi-variant adaptability to network path conditions. So, in order to overcome such drawbacks, an enhanced MP-OLSR-Deficit Round Robin (DRR) combined with QoS metrics delay, and mobility speed of nodes is proposed. To implement proposed system, network simulator NS2 is presented along with result showing packet delivery ratio, end-to-end delay and jitter.

A Polling-Based Traffic-Aware MAC Protocol for Centralized Full-Duplex Wireless Networks

Recent advances in self-interference cancellation techniques have enabled in-band full-duplex (IBFD) transmission, which can double ergodic capacity and reduce end-to-end delay. However, how to solve the channel contention problem in IBFD radios with inter-node interference and asymmetric traffic is still a challenge. This paper presents pFD-MAC, a novel polling-based traffic-aware medium access control (MAC) protocol for centralized full-duplex wireless networks. To solve the channel contention problem and fully utilize the channel resources under full-duplex mode, we first design a novel polling-based transmission mechanism and make comprehensive investigations on the effect of polling profile in full-duplex communication. By characterizing the inter-node interference into a directed non-conflict graph, we study the polling profile generation problem in which our objective is to minimize the packet transmission time. The problem is then theoretically formulated and proved to be NP-hard, which means it cannot be solved in polynomial time. Thus, we develop a heuristic traffic-aware algorithm and apply it to work with the packet transmission procedure in parallel. Full-duplex communication opportunities are highly exploited by parallelly organizing per-node upstream/downstream traffic according to the generated polling profile. Moreover, to achieve fairness without sacrificing throughput, deficit round robin algorithm has been applied with respect to the access time considering concurrent transmission time. Simulation results reveal that our proposed protocol can achieve improved performance in terms of throughput and transmission delay while maintaining fairness, compared with two state-of-the-art centralized MAC protocols.

Constrained max-min fair scheduling of variable-length packet-flows to multiple servers

In this paper, we study a multi-server queuing system wherein each user is constrained to get service only from a specified subset of servers. Fair packet scheduling in such a setting poses novel challenges that we address in this paper. Specifically, we observe that max-min fair allocation of the available resource over different servers (notably bandwidth) in the presence of placement constraints results in different levels of fair service-rates. To achieve the max-min fair service rates, we propose a novel packet scheduler which is inspired by the deficit-round robin (DRR) algorithm. The scheduler allocates tokens to flows in a round-by-round manner, where token allocation to flows at the beginning of each round is weighted max-min fair. So, we have called it multi-server max-min fair DRR (MSMF-DRR). The performance of the MSMF-DRR algorithm in terms of achieving fairness is shown through a worst-case performance analysis. In addition to analytical results, numerical experiments are also carried out to illustrate service isolation and the delay guarantee that are provided by the algorithm. Generally, a scheduler for such a constrained multi-server queuing system can be applicable in many modern data-networking applications, especially in cloud computing wherein virtual machines and/or processes vie for different IT resources distributed over heterogenous servers, while different processes may have preferences over servers owing to their quality-of-service requirements and the heterogeneity of servers.

QoS routing with worst-case delay constraints: Models, algorithms and performance analysis

In a network where weighted fair-queueing schedulers are used at each link, a flow is guaranteed an end-to-end worst-case delays which depends on the rate reserved for it at each link it traverses. Therefore, it is possible to compute resource-constrained paths that meet target delay constraints, and optimize some key performance metrics (e.g., minimize the overall reserved rate, maximize the remaining capacity at bottleneck links, etc.). Despite the large amount of literature that has appeared on weighted fair-queueing schedulers since the mid ’90s, this has so far been done only for a single type of scheduler, probably because the complexity of solving the problem in general appeared forbidding. In this paper, we formulate and solve the optimal path computation and resource allocation problem for a broad category of weighted fair-queueing schedulers, from those emulating a Generalized Processor Sharing fluid server to variants of Deficit Round Robin. We classify schedulers according to their latency expressions, and show that a significant divide exists between those where routing a new flow affects the performance of existing flows, and those for which this do not happen. For the former, explicit admission control constraints are required to ensure that existing flows still meet their deadline afterwards. However, despite this major difference and the differences among categories of schedulers, the problem can always be formulated as a Mixed-Integer Second-Order Cone problem (MI-SOCP), and be solved at optimality in split-second times even in fairly large networks.

A-Duplex: Medium Access Control for Efficient Coexistence Between Full-Duplex and Half-Duplex Communications

As full-duplex wireless communication evolves into a practical technique, it will be built into communication nodes in many application scenarios. However, it is difficult to do so for legacy communication nodes. Thus, full-duplex communication nodes will coexist with half-duplex communication nodes in the same application environment. In this paper, a wireless local area network with a full-duplex access point (AP) and half-duplex clients is studied, and a media access control (MAC) protocol called asymmetrical duplex (A-Duplex) is developed to support efficient coexistence between half-duplex clients and the full-duplex AP. A-Duplex explores packet-alignment-based capture effect to establish dual links between the AP and two different clients. In this way, the capability of a full-duplex AP can be utilized by half-duplex clients, which leads to much improved network throughput. Moreover, to ensure fairness of the MAC protocol, a virtual deficit round-robin algorithm is proposed for the AP to select appropriate half-duplex clients for dual-link setup. A-Duplex does not require any change in the physical layer of half-duplex clients; only an update of MAC driver is necessary. Thus, it is well suited for coexistence between half-duplex clients and a full-duplex AP. Both analysis and simulations are conducted to evaluate performance of A-Duplex. Results show that it improves the throughput by 48% and 188% and reduces the average packet delay by 26% and 22%, as compared to the IEEE 802.11 Distributed Coordination Function with and without RTS/CTS, respectively. Moreover, the throughput remains steady as the number of clients grows. A-Duplex also maintains a high level of fairness.

On Technology Transfer

I agree with the editors of this series that technology transfer from academia occurs in a variety of ways with different tradeoffs. I describe, to the best of my recollection, some of my experiences with ideas that originated or were described in academic papers, from Deficit Round Robin to Conga.

Predictively booting nodes to minimize performance degradation of a power-aware web cluster

With the ever increasing trend of dynamic and static content web, clusters have been widely used for large-scale web servers to improve the system scalability. Dynamically switching the cluster nodes between different power states is one effective approach to save the energy in such clusters. Many research efforts have been invested in designing power-aware clusters by using this method. However, booting a cluster node from a low-power state to an active state takes a certain amount of time that depends on different configurations. This process incurs significant performance degradation. The existing work normally trades a certain amount of performance degradation for energy saving. This paper proposes a hybrid method to predict the number of requests per booting time of the web workloads. A power-aware web cluster scheduler is designed to divide the cluster nodes into an active group and a low-power group. The scheduler attempts to minimize the active group and maximize the low-power group, and boot the cluster nodes in the low-power group in advance to minimize/eliminate performance degradation by leveraging the prediction scheme. Furthermore, this paper integrates the power awareness into the conventional load balancers including Least Connections, Deficit Round Robin, and Skew. Comprehensive experiments are performed to explore the potential opportunities to minimize/eliminate the performance degradation of the power-aware web cluster.

Reducing the execution time of fair-queueing packet schedulers

Deficit Round Robin (DRR) is probably the most scalable fair-queueing packet scheduler. Unfortunately, it suffers from high delay and jitter with respect to a perfectly fair (and smooth) service. Schedulers providing much better service guarantees exist, but have a higher computational cost.In this paper we deal with this issue by proposing a modification scheme for reducing the amortized execution time of the latter, more accurate schedulers. Modified schedulers preserve guarantees close to the original ones, and can also handle seamlessly both leaves and internal nodes in a hierarchical setting.We also present Quick Fair Queueing Plus (QFQ+), a fast fair-queueing scheduler that we defined using this scheme, and that is now in mainline Linux. On one hand, QFQ+ guarantees near-optimal worst-case deviation with respect to a perfectly fair service. On the other hand, with QFQ+, the time and energy needed to process packets are close to those needed with DRR, and may be even lower than with DRR exactly in the scenarios where the better service properties of QFQ+ make a difference.