SDN-Enabled Time-Sensitive Network Reconfiguration
Abstract
We propose to develop a general model and solution to determine how network routing can be reconfigured quickly without incurring transient congestion. Assuming both initial and target configurations are congestion-free, it is known that transient congestion may still occur during the reconfiguration process if links contain a mix of traffic flows following ld and new routing rules, resulting from variation of switch reaction time and propagation delay differences among paths. We consider these factors by explicitly incorporating timing uncertainty intervals into the model. The model leads to an optimization problem whose solution represents a fast (in terms of actual physical time) congestion-free routing reconfiguration. Our formulationnaturally reduces to existing work of finding minimal number of algorithmic update steps when the timing uncertainty intervals are very large, meaning we have little prior knowledge about them. The optimization problem is shown to be a Mixed Integer Linear Program (MILP) with a polynomial-size constraint set, and is proved to be NP-hard. We then further introduce an approximation algorithm with performance guarantee to solve the problem efficiently. In particular, it is demonstrated that timing information can possibly accelerate the update process, even if more steps are involved.Concretely, we propose to do the following three tasks. We need to characterize the optimal solution itself. Thisincludes both optimality condition and complexity of the solution. We are also very interested in tighter approximation algorithms as well as other fast effective heuristics. We then plan to develop model of network traffic dynamics. The model would include two independent variables: user input rate (demand) and router split ratio. There are two main features of this model: careful description of the varying queueing delay and spatial heterogeneity of a fol w s rate due to fferent propagation delay. Both are necessary details (though typically ignored in existing literature) for predicting fast network dynamics. Finally, we hope investigate and compare various network architecture choices such as, for example, distributed vs centralized routing control.Besides its broad intellectual value, the potential contribution of this project for DoD is that by realizing fast routing reconfiguration, it will provide a key technology to operate a robust and highly efficient communication network in a fast varying environment, which is typical for what Navy and Marine Corps networks face. For example, this technology provides the ability to reprioritize traffic dynamically within a network to maximize mission responsiveness to changing battlefield needs, hence enabling a network command authority to choose network traffic prioritization policies on they " instead of being saddled with one-size-fits-all configurations for an entire mission. Furthermore, it is also a critical capability to guarantee delivery of traffic across a multi-hop (wireline or wireless) network within performance (such as latency) requirement in a dynamic environment. Finally, since this technology is independent of the actual performance metric to be used, the fast reconfiguration can be performed based on different (even multiple) parters such as throughput, latency, priority, jitter, etc.).
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 05, 2017
- Source ID
- N000141712419
Entities
People
- A. Kevin Tang
Organizations
- Cornell University
- Office of Naval Research
- United States Navy