David Lastine

Iowa State University, Ames, Iowa, United States

Are you David Lastine?

Claim your profile

Publications (6)0 Total impact

  • Source
    David Lastine · Suresh Sankaran · Arun K Somani
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper we propose a new efficient fault tolerant mul-tipoint routing algorithm for optical networks. The routing for a multi-point request is accomplished by finding a bidirectional cycle simple or nonsimple including all nodes that are participating in the multipoint session. Each link can be used only once. Use of a cycle ensures that a single link (or node in case of simple cycle) failure does not interrupt the session except the failed node if it was part of the multipoint session. De-termining the smallest cycle with a given set of Multi-point (MP) nodes is a NP-Complete problem. Therefore, we explore heuristic algorithms to determine an appropriate cycle to route multipoint connections. We allow non-simple cycles to route requests as they use fewer resources than simple cycles in some cases. We also provide an ILP formulation for routing multipoint request and compare its results with the output of our best heuristic algorithm. On Arpanet for over 80% of the time, our best heuristic is able to find a cycle that is within 1.2 times that of the optimal.
    Preview · Article · Jan 2012
  • Arun K. Somani · David Lastine · Suresh Sankaran
    [Show abstract] [Hide abstract]
    ABSTRACT: A cycle could support a SONET ring, a reliable multicast, a multipoint to multipoint traffic request, or service as a dependable connection, backup paths for requested number of pairwise connection or be used as a p-cycle. In this paper we develop a novel and efficient heuristic to find a cycle in a mesh network that includes a specified set of nodes. We allow links to be used only once in a cycle whereas a node may appear multiple times. This cycle, called a non-simple cycle, reduces the total number of links required to form the cycle. The simulation results show that the proposed heuristic outperforms other cycle finding heuristic algorithms like Optimized Collapsed Ring (OCR) and Enumeration in terms of percent blocking.
    No preview · Conference Paper · Dec 2011
  • Arun K. Somani · David Lastine
    [Show abstract] [Hide abstract]
    ABSTRACT: Light-trails based solutions have been proposed and demonstrated as a means of traffic grooming and optical multicasting in a LAN/MAN, where multiple nodes use time division multiple access on a unidirectional optical bus. When compared to light paths or having nodes relaying traffic using optical-electronic-optical conversion there are advantages and disadvantages to light-trails in terms of bandwidth, hardware requirements and latency. Given that a light-trail of a specific length has been identified, we develop an approach to increase its capacity utilization. In particular, we show that splitting a longer light-trail in shorter segments results into more effective and efficient utilization of bandwidth. However, we do not believe that splitting a light-trail into segments of lengths one is preferable as it will increase the overall delay.
    No preview · Conference Paper · Jun 2010
  • David Lastine · Arun K. Somani
    [Show abstract] [Hide abstract]
    ABSTRACT: Light trails, a fairly recent optical networking architecture for LAN and MAN, and possibly WAN applications, can support both optical unicast and multicast. In this paper we consider routing both types of traffic in a survivable light trail network with multiple wavelengths. The algorithm supports multi-hop routing where traffic may use up to two light trails to reach its destination. An ILP is developed to help plan traffic placement in such a network. Our results show that considering the multicast nature of traffic may be necessary when planning optical routing and protection.
    No preview · Article · Dec 2008
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Energy management is the key issue in the design and operation of wireless network applications like sensor networks, pervasive computing and ubiquitous computing where the network is primarily driven by battery-powered embedded devices. This paper studies network coding as an energy minimization technique. Network coding reduces the energy consumption by minimizing the number of transmissions required to communicate a given amount of information across the network. However, aggressive application of network coding adversely affects the network lifetime. We illustrate this trade off in this paper, and show that the existing throughput based network coding approaches cannot be applied to energy-constrained networks. Specifically, we address the following routing problem. Given a set of traffic demands the goal is to route the demands across the network with the objective of minimizing the total energy consumption while providing guarantees on the lifetime of individual nodes. This paper studies multi-path variation of the above routing problem. We present analytical formulations to solve the problem optimally. Evaluation results indicate that the proposed solution is 35% more energy efficient than no-network- coding solution while still meeting required lifetime constraints.
    Preview · Conference Paper · Aug 2008
  • David Lastine · A.K. Somani
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper considers link based protection where backup capacity is preconfigured into various patterns. When the amount of backup capacity available on a link is constrained, the choice of patterns under consideration can determine if a given network is protectable. This paper gives an ILP that can be used in calculating preconfigurations, even when p-Cycles with p-lines or non-simple p-Cycles alone may not be able to. This ILP includes constraints that will cause backup capacity to be distributed in a somewhat even fashion when possible. When considering networks with dynamic traffic, this may offer the advantage of lower blocking probability over schemes that concentrate backup capacity on a few links. The advantage is demonstrated by a simulation of NSFnet comparing blocking probability of a distribution computed by this ILP vs. a Hamiltonian p-Cycle which is a capacity optimal way to protect the network.
    No preview · Conference Paper · Jun 2008