Changeset 2784

Show
Ignore:
Timestamp:
10/30/09 15:07:40 (4 weeks ago)
Author:
stevedh
Message:
  • fix refs
  • more talking
Location:
docs/Lowthane/ipsn10
Files:
5 modified

Legend:

Unmodified
Added
Removed

  • docs/Lowthane/ipsn10/design.tex

    r2778 r2784  
    7575provides a locally-maintained mechanism for optimizing the collection 
    7676workload.  Therefore, the first primitive {\lowthane} provides is a reliable 
    77 route towards a \controller: a {\emph{Default Route Table}}~\footnote{The semantics of the name become apparent shortly.  In brief, these routes serve as the default backup for cache misses or route failures.}, made up of a list 
     77route towards a \controller: a {\emph{Default Route Table}}\footnote{The semantics of the name become apparent shortly.  In brief, these routes serve as the default backup for cache misses or route failures.}, made up of a list 
    7878of entries, each containing the link-layer address of a node in the direction of a 
    7979\controller.  Each solid arrow in Figure \ref{fig:arch} corresponds to an 
    … …  
    318318a simple mechanism for installing additional {\controller}s. 
    319319 
    320 All topology updates from nodes are sent to a well-known IP multicast group~\footnote{Since node's have no Flow Table entry for this address, packets are sent along the default route, and the {\controller} that receives the packet forwards it to the multicast group.} 
     320All topology updates from nodes are sent to a well-known IP multicast group\footnote{Since node's have no Flow Table entry for this address, packets are sent along the default route, and the {\controller} that receives the packet forwards it to the multicast group.} 
    321321consisting of all {\controller}s.  By listening to routing updates from this group, all controllers 
    322322build the same topology database.  Some other protocol is used to provide 

  • docs/Lowthane/ipsn10/evaluation.tex

    r2778 r2784  
    22 
    33Hui has presented a complete IPv6 architecture for L2Ns which included a 
    4 routing layer \ref{hui}, but there has been no systematic evalutation of the routing 
     4routing layer \cite{hui}, but there has been no systematic evalutation of the routing 
    55performance in the literature.  Since we embrace the IP architecture proposed 
    66by Hui and extend his routing mechanisms, this work is the first in-depth 
    … …  
    135135CTP, the leading open-source protocol.  For point-to-point routing, we compare 
    136136to TinyAODV, a compact implementation of the AODV 
    137 on-demand distance vector routing protocol \cite{tinyaodv}.   
     137on-demand distance vector routing protocol \cite{aodv}.   
    138138Conceptually, the combination of CTP and AODV is very 
    139139similar to 802.11s, which is perhaps the existing architecture closest to our 
    … …  
    249249%In {\lowthane}, the global topology view maintained by the {\controller} enables it to route packets to destinations within the subnet, and install efficient routes in the network for active flows.  However, the optimality of the routes is determined by the accuracy and completeness of its global view, which is a function of the topology reports received from the nodes.  Given the constrained nature of L2Ns, quantifying the tradeoff between optimality and control overhead becomes critical, particularly when the network is first being initialized and the global topology view is empty. 
    250250We begin by examining the average degree of nodes in the global topology view 
    251 at the {\controller}.  T% he intuition is that a larger average degree indicates 
     251at the {\controller}.  % The intuition is that a larger average degree indicates 
    252252% that an increasing portion of the real topology has been captured, and so that 
    253253% links which reduce stretch are more likely to have been discovered.   

  • docs/Lowthane/ipsn10/intro.tex

    r2778 r2784  
    1717% the problem, or solved the problem in another domain with limited 
    1818% applicability to sensornets.   
    19 Among existing work, collection protocols~\cite{mintroute,ctp} 
     19Among existing work, collection protocols 
    2020provide reliable collection to a sink, but don't provide for 
    21 unicast communication.  Point-to-point protocols~\cite{s4-nsdi, tinyaodv, bvr} 
     21unicast communication\cite{mintroute,ctp}.  Point-to-point protocols 
    2222allow for communication between arbitrary pairs of nodes, yet make no 
    23 provisions to optimize common-case collection.  Outside of sensornets, general 
    24 networking platforms such as Ethane~\cite{ethane} provide compelling 
     23provisions to optimize common-case collection \cite{s4-nsdi, aodv, ref:bvr}.   
     24Outside of sensornets, general 
     25networking platforms such as Ethane provide compelling 
    2526solutions, yet are based on assumptions that make porting anything beyond the 
    26 high-level design to sensornets impractical. 
     27high-level design to sensornets impractical \cite{ethane}. 
    2728 
    28 {\lowthane} is not a highly novel new routing protocol; rather our main 
    29 contribution is a seemingly intuitive routing architecture designed from 
    30 scratch that simply works.  We leverage the asymmetrical resource allocation 
     29Our main 
     30contribution is an analysis of a seemingly intuitive routing architecture that simply works.   
     31We leverage the asymmetrical resource allocation 
    3132typical of sensornets to push complexity to {\controller}s at the edge and 
    3233minimize state and functionality in the network.  Our design revolves around 
    3334controlling the tradeoff between state and stretch, and minimizing control 
    34 traffic through a focus on data-driven design.  We embrace the open-standards 
    35 approach to system design, and our solution, {\lowthane}, is in fact the 
    36 default routing protocol for blip, a widely-used IPv6 network layer for 
    37 sensornets. 
     35traffic through a focus on data-driven design.%   We embrace the open-standards 
     36% approach to system design, and our solution, {\lowthane}, is in fact the 
     37% default routing protocol for blip, a widely-used IPv6 network layer for 
     38% sensornets. 
     39As a result, we show that the protocol enables canonical collection workloads, 
     40while adding efficient point-to-point routing with a minimum of additional 
     41cost.  This design is a step in the direction towards more hetrogeneous 
     42networks where sensors, actuators, and controlls can directly communicate 
     43while standard tools are used for debugging. 
    3844 
    39 Beyond presenting {\lowthane}, we provide an analysis from real-world 
    40 experiments that seeks to answer the questions, how well does this general 
     45Our analysis of real-world 
     46experiments seeks to answer the questions, how well does this general 
    4147architecture compare to specialized point solutions, and how robust is it?   
    42 We compare to other protocols which solve a piece of the problem, 
     48We compare to other protocols which each solve a piece of the problem, 
    4349CTP for collection traffic, and TinyAODV for 
    4450point-to-point traffic.  For robustness, 

  • docs/Lowthane/ipsn10/lowthane.bib

    r2771 r2784  
    20282028} 
    20292029 
    2030 @inproceedings{4bit, 
    2031         author    = {Rodrigo Fonseca and Omprakash Gnawali and Kyle Jamieson and Philip Levis}, 
    2032         title     = {Four-Bit Wireless Link Estimation}, 
    2033         booktitle = {Sixth Workshop on Hot Topics in Networks (HotNets VI)}, 
    2034         year      = {2007} 
    2035 } 
    2036  
    20372030@inproceedings{prabalbatch, 
    20382031        author    = {Prabal Dutta and David Culler and Scott Shenker}, 
    … …  
    20512044} 
    20522045 
    2053 @inproceedings{ref:hydrowatch, 
     2046@inproceedings{hydrowatch, 
    20542047      title={Design, Modeling, and Capacity Planning for Micro-Solar Power Sensor Networks}, 
    20552048      author={Jay Taneja and Jaein Jeong and David Culler}, 
    … …  
    29282921 } 
    29292922 
    2930 @article{WASW05, 
    2931   author = "Geoffrey Werner-Allen and Patrick Swieskowski and Matt Welsh", 
    2932   title = "MoteLab: A Wireless Sensor Network Testbed", 
    2933   journal = "IEEE SPOTS", 
    2934   year = 2005, 
    2935 } 
    2936  
    29372923@article{BRY+04, 
    29382924  author = "Maxim A. Batalin and Mohammad Rahimi and Yan Yu and Duo Liu and Aman Kansal and Gaurav S. Sukhatme and William J. Kaiser and Mark Hansen and Gregory J. Pottie and Mani Srivastava and Deborah Estrin", 
    … …  
    35183504  journal = {SIGCOMM Comput. Commun. Rev.}, 
    35193505  volume = {35}, 
    3520   % number = {5}, 
    35213506  year = {2005}, 
    35223507  issn = {0146-4833}, 

  • docs/Lowthane/ipsn10/overview.tex

    r2778 r2784  
    44\caption{An example of Hydro point-to-point routing.  To send the 
    55  first packet (solid red line) from node 1 to node 6, it is routed 
    6   along the default path to the closest controller (node 1). The 
    7   controller source routes the packet to destination and installs a 
     6  along the default path to the closest {\controller} (node 1). The 
     7  {\controller} source routes the packet to destination and installs a 
    88  source route at node 4 (dotted green lines).  Subsequent traffic 
    99  takes a direct path between nodes 4 and 6 (dashed blue line).} 
    … …  
    2424protocols, such as data-driven link estimation~\cite{ctp}, density 
    2525sensitive (trickelized) neighbor discovery messages~\cite{trickle}, 
    26 and careful ranking of multiple potential next hops~\cite{4bitle}. 
     26and careful ranking of multiple potential next hops~\cite{4bit}. 
    2727These techniques enable simple yet robust multipoint-to-point routing. 
    2828{\lowthane} uses this functionality to route packets from within the