A Declarative and Expressive Approach to Control Forwarding Paths in Carrier-Grade Networks

Abstract

SDN simplifies network management by relying on declarativity (high-level interface) and expressiveness (network flexibility). We propose a solution to support those features while preserving high robustness and scalability as needed in carrier-grade networks. Our solution is based on (i) a two-layer architecture separating connectivity and optimization tasks; and (ii) a centralized optimizer called DEFO, which translates high-level goals expressed almost in natural language into compliant network configurations. Our evaluation on real and synthetic topologies shows that DEFO improves the state of the art by (i) achieving better trade-offs for classic goals covered by previous works, (ii) supporting a larger set of goals (refined traffic engineering and service chaining), and (iii) optimizing large ISP networks in few seconds. We also quantify the gains of our implementation, running Segment Routing on top of IS-IS, over possible alternatives (RSVP-TE and OpenFlow). ABSTRACT SDN simplifies network management by relying on declarativity (high-level interface) and expressiveness (network flexibility). We propose a solution to support those features while preserving high robustness and scalability as needed in carrier-grade networks. Our solution is based on (i) a two-layer architecture separating connectivity and optimization tasks; and (ii) a centralized optimizer called DEFO, which translates high-level goals expressed almost in natural language into compliant network configurations. Our evaluation on real and synthetic topologies shows that DEFO improves the state of the art by (i) achieving better trade-offs for classic goals covered by previous works, (ii) supporting a larger set of goals (refined traffic engineering and service chaining), and (iii) optimizing large ISP networks in few seconds. We also quantify the gains of our implementation, running Segment Routing on top of IS-IS, over possible alternatives (RSVP-TE and OpenFlow).

Incremental Deployment of SDN in Hybrid Enterprise and ISP Networks

“…Table 2 summarizes the real ISP and enterprise network topologies we evaluated and the size of end-toend traffic demands placed into each network. We evaluated six real-world ISP topologies, labeled ISP1 through ISP6, using traffic demands [16] that are publicly available at [2]. The enterprise network topology, ENTR, is parsed from router configurations of a real campus network [22].…”

“…Besides, with injected "lies", Fibbing could lead to debugging issues and incorrect operation. DEFO [16] leverages segment routing to control routing paths for carrier-grade traffic engineering, but shares some similar limitations as in Fibbing. Besides, its constraint programming based middle-point selection largely focuses on static traffic matrices, while our proposed TE module can load balance dynamic traffic demands.…”

Let there be Chaining: How to Augment your IGP to Chain your Services

“…Several limitations of this approach have been identified. First, these rules grow with the number of flows, policies and chains' size, while forwarding state on network appliance is limited by costly memory [9]. Second, they grow in complexity when a vSF make a hard to handle change in network headers (e.g., Network Address Translation service) [7], [8], [10].…”

The Segment Routing Architecture

“…By simulating the centralized controller approach on many operator use cases, we observed that the number of SR TE tunnels to be injected to comply with SLA constraints, and re-optimization of such tunnels during failures, led to much less signaling overhead when compared to the distributed RSVP-TE based approach. [13] reports reductions of up to an order of magnitude in the state maintained in routers for 10 different topologies by using SR instead of RSVP-TE. Furthermore, studies have shown that only a few segments are required to obtain optimal traffic distributions: [13] obtains traffic distribution close to the ones of RSVP-TE using only 2 labels.…”

“…[13] reports reductions of up to an order of magnitude in the state maintained in routers for 10 different topologies by using SR instead of RSVP-TE. Furthermore, studies have shown that only a few segments are required to obtain optimal traffic distributions: [13] obtains traffic distribution close to the ones of RSVP-TE using only 2 labels. Independently, [14] also informs that two segments were required to achieve solutions close to the theoretically optimal ones.…”