Scalable Software-Dened Networking

The networking industry is expected to undergo significant transformation with the rise of emerging technologies such as Software Defined Networking (SDN). SDN overcomes the limitations of traditional networks such as complexity, inconsistent policies, and vendor dependency. It improves the management of large networks, offers greater flexibility, and enhances user experience. However, in reality the transition from the legacy network to an OpenFlow-enabled network does not happen overnight. Due to multi-dimensional challenges such as technical, financial and business challenges. Therefore, the network is likely to transition through a hybrid deployment with both legacy and OpenFlow switches and with the same high-level policy implemented through different low-level mechanisms.

A hybrid deployment of SDN can be one of the plausible intermediate paths, primarily because it provides an environment where both legacy and SDN nodes can work together.

In this paper, I present a comprehensive survey of hybrid switch Networking models, techniques, inter-paradigm coexistence and interaction mechanisms. Firstly, I delineate an overview of SDN and hybrid switches and consequently I discuss the de?nition, bene?ts and limitations of hybrid switching. Further, I categorize the di?erent models under various headings, which integrates traditional switching and SDN switching for the purpose of achieving both scalability and optimal performance

1 INTRODUCTION

The capacity of the current Internet is rapidly becoming insu?cient to cater to the large volumes of tra?c patterns delivered by the new services and modalities, which is generated due to a large number of users, sensors and applications. Existing networks built with multiple tiers of static Ethernet switches arranged in a tree structure are ill-suited for the dynamic computing and storage needs of today’s and future enterprise hyper-scale data centers, campuses, and carrier environments. Instead, new networking infrastructures are desired that will provide high performance, energy e?ciency, and reliability. Moreover, they should improve the network speedup, scalability and robustness with the effective creation and delivery of versatile digital services that provide stringent quality of service (QoS) guarantees. 

Meeting these requirements is impossible with existing network equipment due to their limited capabilities. Additionally, today’s protocols tend to be de?ned in isolation and are meant to solve a speci?c problem without the bene?t of any fundamental abstractions. In addition, to implement network-wide policies and to support any new services, managers today have to con?gure thousands of network devices and protocols, which makes it di?cult to apply a consistent set of QoS, security, and other policies. Networks become vastly more complex with the addition of thousands of network devices that must be con?gured and managed. These devices have their control and forwarding logic parts both integrated in monolithic, closed, and mainframe-like boxes. Consequently, only a small number of external interfaces are standardized but all of their internal ?exibility is hidden. The internals di?er from vendor to vendor, with no open software platform to experiment with new ideas.

A lack of standard open interfaces limits the ability of network operators to tailor the networks to their individual environments and to improve either their hardware or software. Hence, there is a need for a new network equipment architecture that decouples the forwarding and control planes of the routers to dynamically associate forwarding elements and control elements that is Software Defined Networking (SDN).

Software-De?ned Networking (SDN) has emerged as the network architecture where the control plane logic is decoupled from the forwarding plane. SDN is a new approach for network programmability, which refers to the ability to control, change, and manage network behavior dynamically through software via open interfaces in contrast to relying on closed boxes and proprietary de?ned interfaces. The SDN framework enables centralized control of data path elements independently of the network technology used to connect these devices that can originate from di?erent vendors. The centralized control embeds all the intelligence and maintains a network wide view of the data path elements and links that connect them. This centralized up-to-date view makes the controller suitable to perform network management functions while allowing easy modi?cations to the networking functions through the centralized control plane.

However, due to a variety of reasons, such as budget constraints and fear of downtime, many organizations are reluctant to fully deploy SDN. Partially deploying SDN through the placement of a limited number of SDN devices among legacy (traditional) network devices, forms a so called hybrid switch network. While hybrid switch networks provide many of the bene?ts of SDN and have a wide range of applications.

Software-defined networking (SDN) has been considered as a break-through technology for the next-generation Internet. It enables fine-grained flow control that can make networks more flexible and programmable. However, this might lead to scalability issues due to the possible flow state explosion in SDN switches. SDN-based source routing can reduce the volume of flow-tables significantly by encoding the path information into packet headers.

2 Types of switches

2.1 Traditional Switching

A traditional Ethernet switch uses a switch table to learn reachability information from the packets (or data frames in layer-2 terminology) that it receives. When a switch receives a frame from a port, it learns that the source MAC address in the frame header can be reached from that port. This information is stored in the switch table where each entry contains an MAC address and a port number. If a switch receives a frame whose destination MAC address is in the switch table, the switch will forward the packet to the corresponding port. Otherwise, it will forward the frame to all ports except for the port from which the frame is received, generating a broadcast. For two-way communication between two hosts, broadcast will happen only once because the ?rst exchange between the hosts will let all switches along the communication path learn how to reach them. To achieve high throughput, the switch table is often implemented as a hash table in SRAM.

2.2 Openflow/SDN Switches

In a software defined networking system, canned processes are used to provision the network. For instance, users should be able to program the network when they want to build a tap, instead of building a network tap using an appliance. SDN makes the network programmable by separating the control plane (telling the network what goes where) from the data plane (sending packets to specific destinations). It relies on switches that can be programmed through an SDN controller using an industry standard control protocol, such as OpenFlow. A OpenFlow/SDN switch, when it receives a packet, that it does not have a flow for will contact a SDN controller and ask what must it do with this packet. The controller can then download a flow to the switch, possibly including some packet manipulation. Once the flow is downloaded to the switch it will switch similar packets at wire-speed having a central server that knows the network layout and can make all the switching decisions and build the paths gives us new capabilities.

3 Hybrid Switches

Hybrid SDN refers to a networking architecture where both centralized and decentralized paradigms coexist and communicate together to di?erent degrees to con?gure, control, change, and manage network behavior for optimizing network performance and user experience. For example, traditionally switches with their distributed algorithms such as IGP (Interior Gateway Protocol) try to control overall tra?c routing whereas, in SDN, the controller routes tra?c based on the global view. If these are combined, say a part of tra?c is under traditional control and the remaining under the SDN controller, we get a hybrid switching architecture.

The main hybrid switching pillars are:-

Coexistence: As the name suggests, this implies a heterogeneity in the infrastructure either in the data plane or the control plane or both. Components of both SDN and the legacy paradigm stand together in the network, although they may or may not interact together. Strategic placement of SDN nodes gives rise to various incentives for a transition. Di?erent placement strategies form the attributes of this pillar.

1. Coexistence at the Data plane only, i.e., SDN and legacy nodes together exist in the network, but managed by the distributed control plane of the legacy paradigm only. Although this setup is possible, it provides little bene?t as managing SDN nodes with legacy control o?ers no advantage.
2. Coexistence at the Control plane only, i.e., centralized SDN control and decentralized legacy control both prevail in the network. For example, [2] propose a routing controller which provides a consistent assignment of routes for external tra?c with the help of a global topology view. The routers pull the routes for external tra?c from this controller, but the controller does not interact with the legacy control. This is coexistence without any communication.
3. Coexistence at both data and control planes. Here, the network contains both SDN and legacy components, both in the control plane (SDN controller and distributed legacy control) as well as the data plane (SDN and legacy nodes). For example, Telekinesis [6] introduces a mechanism to control the routing in legacy devices with OpenFlow protocol in a hybrid network.

Communication: Communication conveys the idea of inter-paradigm integration with mutual understanding and sharing & distributing of functionality among fundamentally heterogeneous components of the network. SDN and legacy components not only co-exist but also interact with each other and understand the interfaces and the protocols of each other to enhance each other. This involves a number of techniques like protocol translation, SDN nodes stealing the legacy (control) packet away to the controller, parsing packets, packet injection from the controller to the network and so on. For example, the SDN controller may help with the global topological view, whereas the local legacy control in the node may take fast local decisions even when the node-SDN controller link is congested. Sometimes communication is crucial. For example, in a network having a legacy and SDN switches, it is not possible to have a loop-free Layer-2 network unless both understand STP (spanning tree protocol) protocol. Hence, this is an important pillar. Communication can be realized at di?erent planes.

Crossbreeding: Crossbreeding involves intermixing di?erent paradigms whose complementary attributes enhance the hybrid network. Here, crossbreeding indicates the degree of hybridization in terms of the following attributes that dictates architectural trade-o?s. For example, there can be a trade-o? between the number of features of a legacy protocol the SDN controller may parse and interpret versus the performance of the controller. Similarly, as the number of SDN nodes increase in the data plane, more tra?c comes within SDN control, although this increases the budget for the organization .Therefore, these often act as parameters for a particular implementation to be chosen by an organization for deployment.

3.1 Bene?ts Hybrid Switching

Overview of the speci?c advantages of the hybrid Switching are:-

1. Hybrid switching enables SDN-speci?c features (such as centralized control of the network) coupled with bene?ts of legacy (such as low deployment costs and time tested-maturity). Therefore, it can give the best of both. For example, in a tra?c class-based hybridization model, the policy expression at a high level becomes easy. Hybrid switching provides the feature to fallback to time-tested legacy mechanisms in case of SDN controller failure, which is not available in pure SDN paradigm.
2. SDN network deployment is ?nancially very costly. To replace all the existing legacy devices by SDN devices, large budget amounts are required to purchase SDN devices. After full deployment of SDN, i.e., after creating a pure SDN network, additional budget amounts are required to train the operators to design, con?gure, and operate the SDN network. Hybrid Switching networks ease these budget concerns.
3. There are areas where a combination of centralized and decentralized mechanisms function well. Update or installation of a large number of rules in the devices centrally could be a problem in pure SDN (due to control channel clogging, congestion, the processing capacity of controller etc.). Using both central and distributed control in the same environment, we can overcome this problem. If the communication with the controller is congested or the controller is unable to respond due to lots of loads, the switch can use distributed legacy routing mechanisms in the meantime to route crucial packets. By providing a central control over critical tra?c only, overhead on the controller is reduced and the controller’s scalability can be increased. On the contrary, in pure SDN there has been a lot of work going on to realize a hierarchical model of controllers to guarantee centralization with scalability for large networks
4. Architectural tradeo?s can be tuned to cater to the needs of an organization. For example, based on whether the organization wants to initially incentivize and accommodate the premium users or to enhance telecom billing, there can be di?erent proportion in which the tra?c can be controlled either by SDN or non-SDN paradigm. This can be tuned based on the speci?c needs of the organization.
5. There are economic and business bene?ts like gradual investment, building the con?dence of network operators and end users.
6. SDN provides ?ne-grained control for data traf?c ?ows. If ?ne-grained control is only required for a small portion of the network, then a hybrid SDN network can be implemented by executing SDN for that small portion of network requiring ?ne-grained control, while the rest of the network uses traditional networking.

4. Integration of SDN Switching With Traditional Switching/Routing

We ?rst consider the integration of SDN switching with traditional routing. When a switch receives a packet, it matches the packet against both the SDN forwarding table and the traditional routing table. As long as the forwarding table has a matching entry, it takes the precedence and the packet will be forwarded accordingly. If the packet belongs to a new ?ow and the forwarding table does not have a matching entry, there are two path selection strategies.

Traditional Path First (TPF): New ?ows will take the traditional paths by default without causing any immediate overhead to the controller. For a packet from a new ?ow, without a match in the forwarding table, the switch will handle the packet according to the routing table, which will always give a matching entry, meaning that it can scale to an arbitrary number of ?ows. New ?ows will not automatically generate requests to the controller for path selection, in contrast to what today’s SDN switches do. This property helps reduce the controller’s communication/computation burden and avoid a potential performance bottleneck in the system. While all new ?ows follow the traditional paths by default, the switches will monitor their ?ows, identify the large ones, and estimate their sizes. Periodically they will send the information of the identi?ed large ?ows to the controller, which performs global optimization to improve network performance by rerouting some or all of the large ?ows via optimal SDN paths, subject to the size constraint of the forwarding tables at the switches. The formulation of the optimization is dependent on the user-speci?ed performance and management requirements. The controller will then update the switches’ forwarding tables by installing the new paths.

SDN Path First (SPF): New ?ows will take the SDN paths by default. For a packet from a new ?ow, without a match in the forwarding table, as long as the switch’s forwarding table is not over?own, it will forward the packet header to the controller for installing an SDN path. If the forwarding table is full, the switch forwards the packet based on the routing table. Although SPF solves the over?ow problem of forwarding tables, it still faces other problems of SDN switching, per-?ow communication/ computation overhead to the controller (even for small ?ows that contain a few packets themselves) and extra delay to a ?ow’s ?rst packet due to the setup of SDN path. I advocate TPF not only because it avoids these problems but also because batch setup of forwarding paths for a set of ?ows together tend to produce better global optimization than setup of the paths one at a time sequentially. Next, I consider the integration of SDN switching with traditional switching under TPF. When a switch receives a data frame, it matches the frame against both the SDN forwarding table and the traditional switch table. As long as the forwarding table has a matching entry, the frame will be processed based on that entry; otherwise, if the switch table has a matching entry, the frame will be forwarded to the speci?ed output port. If there is no matching entry in either table, the switch will send a request, carrying the frame’s destination MAC address, to the controller, which will establish a tradition path towards the destination and install proper entries in the switch tables along the path. Again, all switches will monitor their ?ows and send the information of large ?ows to the controller, which will perform global optimization periodically by re-routing large ?ows to optimal SDN paths.

5. Different Forms of Hybrid SDN

Hybrid Software De?ned networking (hybrid SDN), i.e., the combination of legacy (pre-SDN) networking principles with SDN networking principles can take on different forms.

The major categories of hybrid SDN are:-Deployment of SDN switches in a legacy network, i.e., among legacy switches, to form a hybrid SDN network and Hybrid SDN switches having both SDN switching and legacy switching functionality.

1. Deployment of SDN Switches in Legacy Network: Hybrid SDN Network: In this form of hybrid SDN, SDN switches are placed in a legacy network, e.g., among legacy IP switches, to form a so-called hybrid SDN network. By forming a hybrid SDN network, old legacy switches can be used to realize SDN-like control and management in a legacy network.
2. Hybrid SDN Switches with Both SDN and Legacy Switch Functionalities: Network switches can be equipped with both SDN and legacy switch functionalities to form hybrid SDN switches.

6. Parsing and Con?guration Translation

This section surveys parsing and con?guration translation based techniques for the control of hybrid switching networks. The parsing and con?guration techniques translate the information from legacy network devices into a form that is understood by the SDN controller. Hand and Keller  have developed a technique for controlling legacy switches and routers through a central so-called ClosedFlow controller that is similar to an SDN controller. The developed ClosedFlow control technique provides SDN like control over legacy switches and routers.

In pure SDN networks, four major tasks are performed:

1. Establishment of a control channel between the controller and SDN devices, 
2. Knowledge of topology for network-wide view at the controller, 
3. Modi?cation of ?ow tables by changing entries and actions at SDN switches, and 

Communication between controller and SDN devices. ClosedFlow performs these tasks to control and manage legacy devices in a hybrid SDN network as follows: 

1. Controller-switch control channel: a minimum routing instance of OSPF is implemented to create a channel between controller and switch. This includes advertisements for loopback management interfaces of the switch, point-to-point connections between switches, and a VLAN for controller communication. 
2. Topology Discovery: The controller discovers the topology through remote logging at each legacy switch. This allows the controller to store the topology state and to learn about link failures.
3. Packet matching and applying actions are achieved through a combination of access control lists (ACL), Route Maps, and interface con?gurations of legacy devices.
4. Handling Packet-In Event: This event is handled in two ways: (a) remote logging on explicit deny, and (b) send the entire packet to the controller.

The proposed approach does not support other network functions, such as load balancing or loop detections. Also, each type of switch requires a speci?c corresponding ClosedFlow mechanism.

7. CONCLUSION

The Software De?ned Networking (SDN) paradigm has moved network management and control to a centralized controller. While SDN promises a wide range of bene?ts, organizations are often reluctant to replace their entire traditional network by an SDN network due to a variety of reasons, including cost constraints. Incrementally deploying a few SDN devices among the legacy devices in a traditional network, creates a so-called hybrid switching network. A hybrid switching network requires only modest investments into SDN devices and may provide control functionalities approaching those of a pure SDN network. Essentially the hybrid switching network approach still utilizes the installed traditional network infrastructure, while providing SDN-like control and management. In this paper, I have presented a novel hybrid switching mechanism, which integrates traditional switching and SDN switching for the purpose of achieving both scalability and optimal performance. Moreover, a hybrid path deployment method has been presented to reduce the required forwarding rules

Referance

• Tilmans, O., Vissicchio, S., Vanbever, L., Rexford, J., 2016. Fibbing in action: Ondemand load-balancing for better video delivery. In: Proceedings of the ACM SIGCOMM 2016 Conference, ACM, pp. 619–620.
• Caesar, M., Caldwell, D., Feamster, N., Rexford, J., Shaikh, A., van der Merwe, J., 2005. Design and implementation of a routing control platform. In: Proceedings of the 2nd conference on Symposium on Networked Systems Design&Implementation-Volume 2, USENIX Association, pp. 15–28.
• Hartert, R., Vissicchio, S., Schaus, P., Bonaventure, O., Fils?ls, C., Telkamp, T., Francois, P., 2015. A declarative and expressive approach to control forwarding paths in carrier-grade networks. In: ACM SIGCOMM Computer Communication Review, Vol. 45, ACM, pp. 15–28.
• Vissicchio, S., Tilmans, O., Vanbever, L., Rexford, J., 2015. Central control over distributed routing. ACM SIGCOMM Comput. Commun. Rev. 45 (4), 43–56.
• Vissicchio, S., Vanbever, L., Rexford, J., 2014c. Sweet little lies: Fake topologies for ?exible routing. In: Proceedings of the 13th ACM Workshop on Hot Topics in Networks, ACM, p. 3.
• Jin, C., Lumezanu, C., Xu, Q., Zhang, Z.-L., Jiang, G., 2015. Telekinesis: Controlling legacy switch routing with OpenFlow in hybrid networks. In: Proceedings of the 1st ACM SIGCOMM Symposium on Software De?ned Networking Research, ACM, p. 20.
• R. Hand and E. Keller, “ClosedFlow: OpenFlow-like control over proprietary devices,” in Proc. ACM Workshop Hot Topics Softw. De?ned Netw., Chicago, IL, USA, 2014, pp. 7–1
• S. Vissicchio, L. Vanbever, and O. Bonaventure, “Opportunities and researchchallengesofhybridsoftwarede?nednetworks,” ACMSIGCOMM Computer Communication Review, vol. 44, no. 2, pp. 70–75, 2014.

Did you like this example?

1004

42