Gregory Kuperman

Analysis and algorithms for partial protection in mesh networks

This paper develops a novel mesh network protection scheme that guarantees a quantifiable minimum grade of service upon a failure within the network using multipath routing. Typically, networks fully guarantee service after a single-link failure, which is often an over-provisioning of resources to maintain essential traffic for the infrequent event of a failure. Our scheme guarantees that a fraction q of each demand remains after any single-link failure, at a fraction of the price of full protection. A linear program is developed to find the minimum-cost capacity allocation to meet both demand and protection requirements. For q ≤(1/2) , an exact algorithmic solution for the minimum-cost routing and capacity allocation is developed using multiple shortest paths. For q>(1/2) , an algorithm is developed based on disjoint path routing that performs, on average, within 1.4% of optimal, and runs four orders of magnitude faster than the minimum-cost solution achieved via the linear program. Moreover, the partial protection strategies developed achieve reductions of up to 83% over traditional full protection schemes.

Providing protection in multi-hop wireless networks

We consider the problem of providing protection against failures in wireless networks subject to interference constraints. Typically, protection in wired networks is provided through the provisioning of backup paths. This approach has not been previously considered in the wireless setting due to the prohibitive cost of backup capacity. However, we show that in the presence of interference, protection can often be provided with no loss in throughput. This is due to the fact that after a failure, links that previously interfered with the failed link can be activated, thus leading to a “recapturing” of some of the lost capacity. We provide both an ILP formulation for the optimal solution, as well as algorithms that perform close to optimal. More importantly, we show that providing protection in a wireless network uses as much as 72% less protection resources as compared to similar protection schemes designed for wired networks, and that in many cases, no additional resources for protection are needed.

Network protection with guaranteed recovery times using recovery domains

We consider the problem of providing network protection that guarantees the maximum amount of time that flow can be interrupted after a failure. This is in contrast to schemes that offer no recovery time guarantees, such as IP rerouting, or the prevalent local recovery scheme of Fast ReRoute, which often over-provisions resources to meet recovery time constraints. To meet these recovery time guarantees, we provide a novel and flexible solution by partitioning the network into failure-independent “recovery domains”, where within each domain, the maximum amount of time to recover from a failure is guaranteed. We show the recovery domain problem to be NP-Hard, and develop an optimal solution in the form of an MILP for both the case when backup capacity can and cannot be shared. This provides protection with guaranteed recovery times using up to 45% less protection resources than local recovery. We demonstrate that the network-wide optimal recovery domain solution can be decomposed into a set of easier to solve subproblems. This allows for the development of flexible and efficient solutions, including an optimal algorithm using Lagrangian relaxation, which simulations show to converge rapidly to an optimal solution. Additionally, an algorithm is developed for when backup sharing is allowed. For dynamic arrivals, this algorithm performs better than the solution that tries to greedily optimize for each incoming demand.

Simulation and modeling of a new medium access control scheme for multi-beam directional networking

In this paper, we analyze a new medium access control (MAC) protocol for a multi-beam directional network via high-fidelity simulation using a real-time emulator. Multi-beam directional systems are a novel approach to networking which leverage recent advances in physical layer technology, allowing formation of multiple simultaneous beams in both transmit and receive. These multiple beams significantly reduce the burden some coordination between the transmitter and receiver and enable an uncoordinated, distributed random access scheme that offers high throughput. This paper is the first to characterize the performance of such systems using real-time simulation tools. In addition to implementing the random access scheme, several MAC features are developed that allow for robust communication, such as location tracking and extrapolating neighbor transmit or receive state. For this paper, we implement our protocol in both simulation and a new Extendable Mobile Ad-hoc Network Emulator (EMANE) model that allows for real-time, high fidelity performance evaluation. Using EMANE allows us to better understand the performance of the newly developed protocols by running real-time applications through the network. We show that our results from the EMANE model and simulator coincide with the theoretical network throughput, which allows for empirical characterization of scenarios in which theoretical results have not been derived. Furthermore, through our work, we stress test the EMANE model and quantify the maximum number of nodes that we can operate in real-time. Through this testing, two bottlenecks are identified: 1) infrastructure issues, where the amount of data passed between the servers is too high, and 2) computation issues, where calculating the interference on the packets becomes too time consuming.

Group-centric networking: addressing information sharing requirements at the tactical edge

In recent years, there has been a large push in the U.S. Department of Defense to move to an all-IP infrastructure, particularly on the tactical edge. IP and associated protocols were designed primarily for wired networks tied to fixed infrastructure. Although extensions to support MANET have received decades of research, in practice, there are few successful implementations. Challenges include handling mobility, managing lossy links, and scaling to large numbers of users. Unfortunately, these are the exact conditions military tactical edge networks must operate within: high mobility, high loss, and large numbers of users. To address the needs and particular challenges of military tactical edge information sharing requirements, we consider a new class of networking approaches called group-centric networks that focuses on dynamic and resilient formation of interest groups. The structure of tactical networks limits the majority of collaboration and network traffic to within a group of users that share a set of common interests (i.e., platoons, 4-ships, etc.). These groups are formed either prior to the mission or on the fly with only a minor amount of traffic flowing outside of these groups. Group-centric networking approaches can help connect users in military tactical edge networks. In addition, we also discuss an instantiation of a group-centric network protocol called Group Centric Networking (GCN), compare GCN against a traditional MANET routing approach on a 90-node Android mobile phone testbed, and discuss implications for tactical edge users.

Quantifying the impact of routing and scheduling on throughput for wireless networks

Current evaluations of DoD tactical networking systems have left the impression of sub-optimal performance without being able to provide either a clear vision of the limits on the performance that could realistically be attained or which layers of the network algorithm are responsible for the suboptimal results. Recent advances in computing the capacity of a large MANET allow us to obtain a practical benchmark for MANET capacity performance evaluation [1]. In this paper, we quantify the impact of commonly used algorithms at the routing and the scheduling layers on the overall network throughput and compare their individual effects on overall network throughput performance. We consider the routing and the scheduling layers separately since practical MANET implementations are likely to use a layered architecture even though the joint routing and scheduling algorithm is known to be optimal. Our data shows that a good scheduling algorithm can provide potentially four times the throughput improvement of a good routing algorithm when inter-user interference conforms to an 802.11 type model.