Routing optimization with path cardinality constraints in a hybrid SDN
Introduction
The emerging Software Defined Networking (SDN) is a novel networking paradigm where the data plane and control plane are decoupled on the SDN switch and SDN controller, respectively [1]. Concretely, SDN switch is a programmable device that is mainly responsible for forwarding network flows according to flow entries from SDN controller. SDN controller is a logically centralized device that supports a global view of the network state by collecting network information from SDN switches, and controls the fine-grained forwarding of network flows by dispatching flow entries to SDN switches. Instead of constantly forwarding flows according to the shortest paths implemented in traditional (or non-SDN) networking systems, SDN can achieve a more flexible, convenient and intelligent network management by customizing the forwarding rules for each flow [1]. The SDN paradigm greatly simplifies and facilitates the network management. Therefore, the related studies on SDN have been attracting worldwide attentions.
Although SDN brings many advantages, in practice, it is of great difficulties to entirely migrate a traditional network to a SDN network in one step, especially for a large-scale traditional network [2], [3]. The reasons mainly include two aspects: economical and technical factors. On one hand, purchasing and deploying SDN devices to upgrade the infrastructure of entire traditional network will inevitably impose a huge capital expenditure and operational burden on Internet Service Providers (ISP). Currently, ISP may be reluctant to implement one-step migration from traditional network to SDN. Moreover, many studies have demonstrated that it is unnecessary to update all legacy devices with SDN devices by considering deployment benefits [3], [4]. On the other hand, the emerging SDN-related hardware and software are relatively immature and lack of adequate evaluations on large-scale networks [5]. The reliability and stability of SDN devices cannot be guaranteed, which means that updating the entire infrastructure of traditional networks with specialized SDN devices may lead to potential security and instability risks.
Due to the above mentioned economical and technical factors, instead of upgrading the entire network, many ISPs begin to incrementally deploy SDN switches in a traditional network, thereby forming a more practical hybrid SDN with partial deployment of SDN switches in a traditional network [3], [6]. Fig. 1 shows a hybrid SDN scenario. As illustrated in Fig. 1, there are legacy routers and SDN devices working in a hybrid network. More specifically, the legacy routers supports only the OSPF protocol. The SDN switches work in a hybrid mode that supports not only the SDN protocol (e.g., openflow) to connect with the SDN controller, but also the OSPF protocol to interact with the legacy routers. Therefore, the legacy routers can only forward the flow through the shortest paths, which is indicated at the routing tables. However, flows in SDN switches can be allowed to be arbitrarily split for multi-paths forwarding, which is indicated by the flow entries from the SDN controller. Therefore, the hybrid SDN architecture not only has the flexibility of a centralized network, but also has the robustness of a distributed network. In this paper, we mainly focus on studying the routing optimization in a hybrid SDN.
In order to exploit the flexibility provided by SDN devices to achieve an efficient routing in a hybrid SDN, many previous studies have focused on designing effective methods to determine the route selection for flows [7], [8]. To realize it, the network operator usually needs SDN controller to indicate the forwarding behavior of SDN switches, which leads to a large number of flow entries installed in SDN switches. Abundant flow entries installed in SDN switches will inevitable largely consume the expensive and power-hungry Ternary Content Addressable Memory (TCAM) resource [9], [10]. Moreover, a large number of flow entries lead to the increment of routes management complexity. Therefore, it is very promising and practical to consider reduction of the used flow entries into routing optimization. To reduce the number of flow entries used in forwarding flows, taking the path cardinality constraints into routing optimization is necessary.
How to optimize routing with path cardinality constraints to minimize Maximum Link Utilization (MLU) of a hybrid SDN poses a great challenge. Here, we start with load balancing as our objective goal, aiming to minimize the MLU to avoid link congestion. Specifically, the fewer paths the flows route on, the fewer flow entries generated, but the worse network performance we obtain. The path cardinality constraints will lead to the degradation of network performance. To make a trade off between flow entries number and network performance, we should determine the most profitable path cardinality for routing flows in a hybrid SDN. With huge solution space, which and how many paths to select out of all permissible paths for each flow makes the routing optimization problem difficult to solve.
To solve this problem, we first formulate the problem of routing optimization with path cardinality constraints as a Mixed Integer Non-Linear Programming (MINLP) problem, which is NP-hard. Then, we propose an approximation algorithm: H-permissible Paths Routing Scheme (HPRS) to route each flow on paths for minimizing MLU of a hybrid SDN. Next, we analyze and prove the approximation ratio of HPRS through theoretical analysis. Finally, through extensive experiments, we demonstrate that HPRS has superior performance in minimizing MLU and in reducing flow entries over other routing algorithms. Our main contributions in the paper are summarized as follows:
- •
We propose to take the path cardinality constraints into consideration when optimize routing in a hybrid SDN scenario.
- •
We formulate the routing optimization problem as a MINLP model, which is proved to be NP-hard. To solve the problem, we propose an approximation algorithm HPRS with an approximation ratio of ( is the link number) to select permissible paths for flows routing.
- •
We evaluate HPRS on different topologies and traffic datasets. Through experiments, we observe that -permissible paths routing can reap the most of benefit. In addition, comparative experiments show the superiority of HPRS over other routing algorithms in minimizing MLU and reducing flow entries.
The reminder of this paper is organized as follows. Section 2 introduces some related works. Section 3 presents the network model of routing optimization problem under the path cardinality constraints, and analyzes the complexity of the problem. Section 4 proposes an approximation algorithm HPRS with a theoretical guarantee of and analyzes the complexity of the algorithm. Section 5 evaluates the performance of the proposed method and reports extensive experimental results. Finally, Section 6 gives a conclusion of this work.
Section snippets
Related work
In this section, we introduce the related works on routing optimization and TCAM usage optimization, respectively.
Routing optimization. For traditional distributed networks, routing optimization algorithms mainly focus on OSPF link weight optimization. Fortz et al. [11] propose an IP-based intra-domain routing optimization solution. They exploit a refined tabu search heuristic algorithm to search for the optimal OSPF weight setting. Similarly, Ericsson et al. [12] propose genetic algorithm to
Network model
In this section, we formulate the problem of routing optimization with path cardinality constraints in a hybrid SDN as a MINLP problem. The related notions are summarized in Table 1.
Given a hybrid network with nodes and edges, and represent the set of nodes and edges, respectively. Each node represents a SDN switch or a legacy router. Each edge represents a link. A network flow is defined as an end-to-end aggregated traffic demands between two nodes. To minimize MLU
Proposed algorithm
To solve the MINLP problem (1) formulated in the previous section, we propose an approximate HPRS algorithm to consider the path cardinality constraints into optimizing the routing of flows in a hybrid SDN. In this section, we first present the detailed description of the proposed HPRS algorithm. Then, we analyze approximation ratio of the proposed HPRS algorithm, and give its upper and lower bounds. Finally, we present the time complexity analysis of the proposed HPRS algorithm.
Evaluation
In order to demonstrate the performance of our proposed HPRS, different topologies and traffic datasets are used in the experimental evaluation. In this section, we first detail the environment setting. Then, we determine the deployment ratio of SDN nodes and path cardinality through extensive experiments. After that, we demonstrate the performance of the proposed HPRS on minimizing MLU under the determined deployment ratio of SDN nodes and path cardinality. Finally, we present the computation
Conclusion
For a hybrid SDN, routing flexibility comes at the expense of the abundant flow entries storage overhead in SDN switches memory. In this paper, considering the limited space of the expensive hardware resource TCAM, we study the problem of routing optimization under the path cardinality constraints. We formulate the problem as a MINLP problem, which is a NP-hard problem and propose an approximation algorithm HPRS with an approximation ratio of . Through extensive simulation experiments,
CRediT authorship contribution statement
Yingya Guo: Conceptualization, Methodology, Writing - original draft. Huan Luo: Methodology, Software, Writing - review & editing. Zhiliang Wang: Formal analysis, Writing - review & editing. Xia Yin: Supervision, Resources. Jianping Wu: Supervision, Resources.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work is partially supported by the Educational Research Project for Young and Middle-aged Teachers of Fujian Education Department under Grant No. JAT190027, Fuzhou University Fund No. GXRC-19062, National Natural Science Foundation of China under Grant No. 62002064.
References (39)
- et al.
Traffic engineering in hybrid SDN networks with multiple traffic matrices
Comput. Netw.
(2017) - et al.
STAR: Preventing flow-table overflow in software-defined networks
Comput. Netw.
(2017) - et al.
Jumpflow: Reducing flow table usage in software-defined networks
Comput. Netw.
(2015) - et al.
A low overhead flow-holding algorithm in software-defined networks
Comput. Netw.
(2017) - et al.
Hybrid SDN networks: A survey of existing approaches
IEEE Commun. Surv. Tutor.
(2018) - et al.
Opportunities and research challenges of hybrid software defined networks
ACM SIGCOMM Comput. Commun. Rev.
(2014) - et al.
Incremental deployment of sdn in hybrid enterprise and isp networks
- et al.
One step at a time: optimizing sdn upgrades in isp networks
IEEE INFOCOM 2017-IEEE Conference on Computer Communications
(2017) - et al.
Incremental deployment and throughput maximization routing for a hybrid SDN
IEEE/ACM Trans. Netw.
(2017) - et al.
Panopticon: reaping the benefits of incremental SDN deployment in enterprise networks
Traffic engineering in software defined networks
Maximizing network utilization in hybrid software-defined networks
The joint optimization of rules allocation and traffic engineering in software defined network
Efficient TCAM rules distribution algorithms in software-defined networking
IEEE Trans. Netw. Serv. Manag.
Internet traffic engineering by optimizing OSPF weights
A genetic algorithm for the weight setting problem in OSPF routing
J. Comb. Optim.
Determining link weight system under various objectives for ospf networks using a Lagrangian relaxation-based approach
IEEE Trans. Netw. Serv. Manag.
Achieving high utilization with software-driven WAN
B4: Experience with a globally-deployed software defined WAN
Cited by (14)
DRL-based routing algorithm with guaranteed loss, latency and bandwidth in SDN networks: Application of online video conferencing
2023, Journal of King Saud University - Computer and Information SciencesPath selection for link failure protection in hybrid SDNs
2022, Future Generation Computer SystemsCitation Excerpt :Moreover, the controller is responsible for distributing the forwarding rules to the network devices to handle the user-initiated traffic. The hybrid SDNs inherit the benefits of pure SDNs and traditional networks, which are superior to traditional networks in terms of traffic control, network performance and dynamic routing [14–17]. Although the hybrid SDNs have a wide range of applications and advantages, their emergence poses some challenges to the network failure protection [18].
Traffic Engineering in Hybrid Software Defined Network via Reinforcement Learning
2021, Journal of Network and Computer ApplicationsCitation Excerpt :To further improve the network performance in a hybrid SDN, Guo et al. (2014) proposed to jointly optimize the OSPF link weights and the splitting ratio of SDN nodes for the minimization of MLU. To reduce the abundant flow entries installed into the SDN switches, the path cardinality constraints were considered into optimizing the network performance in the hybrid SDN (Ren et al., 2020; Guo et al., 2021). To improve the network performance under different TMs in a hybrid SDN, an expected TM was calculated and exploited to reflect the fluctuating TMs over a period of time (Guo et al., 2017b).
Communication Combination Technology and Routing Protocol Algorithm Design Based on Artificial Neural Networks
2023, Proceedings - 2023 International Conference on Power, Electrical Engineering, Electronics and Control, PEEEC 2023