Transcript Data Centric, Position-Based Routing In Space Networks

Data Centric, Position-Based
Routing In Space Networks
Siva Kumar Tanguturi
&
Sanjaya Gajurel
{skt8, sxg125}@eecs.case.edu
4/6/2005
EECS 600 Advanced Network Topics
Agenda

Introduction to Space Networks

Background
Architecture
Implementation
Simulation Experiments
Conclusions
Discussion





Siva kumar Tanguturi & Sanjaya Gajurel
This is Space
Source: Kul Bhasin, Jeff Hayden., Developing Architectures and Technologies for an Evolvable NASA Space Communication
Infrastructure , 22nd AIAA International Communications Satellite Systems Conference, May 2004
Siva kumar Tanguturi & Sanjaya Gajurel
Space Networks
Backbone or Inter-Planetary or Deep Space
Network

Earth-Mars Network

Earth-Orbital Network

Earth-Lagrangian-Relay-Orbital
(Multi-Hop) Network
Orbital Network

Access Network

Inter-spacecraft & Intra-spacecraft
Network

Inter-Orbital
Proximity Network or Surface Network

Sensor Networks

Inter-Surface Elements Networks
(robots, access points, rovers, landers,
balloons etc. communicating each
other)
Human-Robot Networks
Siva kumar Tanguturi &
Sanjaya Gajurel
Communication Problems in Space
Deep Space







Very High and Variable Propagation
Delay
High Link Error Rates or Error-Prone
Links
Blackouts or Intermittent Connectivity
Bandwidth Asymmetry
Very Low Bandwidth/ Limited Link
Capacity
High Power Requirement
Security
Source: http://www.jpl.nasa.gov/history/hires/1997/VLBI.jpg
Siva kumar Tanguturi & Sanjaya Gajurel
Communication Problems in Space
Orbital
 Latency (Intermittent
connection)
 Gravitational Fluctuations
 The Sun’s interference
 Doppler’s effect in Satellite
Radio Signal
 Orbital Debris
 Distributed Computation
Source: Kul Bhasin, Jeff Hayden., Space Internet Architectures
and Technologies for NASA Enterprises.
Siva kumar Tanguturi & Sanjaya Gajurel
http://mrr.nrl.navy.mil/applications.html
Communication Problems in Space
Surface
 Noise & Power issue
 Highly mobile
 Weight, Cost, & Power
 Harsh Environment
 No infrastructure (Ad
Hoc topology)
Source:http://scp.grc.nasa.gov/images/portfolio/pn/pn
%20main.jpg
Siva kumar Tanguturi & Sanjaya Gajurel
Agenda







Introduction to Space Networks
Background
Architecture
Implementation
Simulation Experiments
Conclusions
Discussion
Siva kumar Tanguturi & Sanjaya Gajurel
Problems with existing TCP/IP
protocol suite

The current approaches cannot support the dynamic nature of
the space networks.

They work well only if the nodes and links are fixed and wellknown ahead of time.

They are not intelligent to discover the links as they become
available and use them for routing.
Siva kumar Tanguturi & Sanjaya Gajurel
Effect of Space Environment on TCP

Effect of Error Prone Links:
TCP is designed to handle packet loss by identifying and
retransmitting lost segments assuming the source of all packet loss
is network congestion

Effect of Asymmetric Channels:
TCP rely on feedback in the form of cumulative acknowledgements
from the receiver to ensure reliability. In addition, TCP is ackclocked, relying on the timely arrival of acknowledgements, to
make steady progress and fully utilize the available bandwidth of
the path.
Siva kumar Tanguturi & Sanjaya Gajurel
Effect of Space Environment of TCP

Effect of Limited Link Capacity :
The packet overhead, at least 20 bytes of TCP header per packet, can
consume a sizable share of a limited bandwidth channel.

Intermittent Connectivity :
Even short-term link outages pose a problem for TCP ranging from poor
throughput in best case to an aborted connections in the worst case.

Extremely long and variable Propagation Delays :
For the very long propagation delay in minutes, TCP has to set its
retransmission timer very long to wait for the acknowledgement. This
long delay is not acceptable. Moreover, because of the changing network
topology, TCP can’t estimates
the optimal timeout.
Siva kumar Tanguturi & Sanjaya Gajurel
Effects of Space Environment on
Network Layer

Naming And Addressing:
If the application on a remote planet wished to resolve an
earth-based address, the long round-trip delay to query the
DNS is significant in terms of available communication time.
With the use of secondary DNS on the surface, addresses
updates have to be sent frequently to the secondary DNS that
can consume a large portion of the limited bandwidth of the
space.
Siva kumar Tanguturi & Sanjaya Gajurel
Effects of Space Environment on
Routing

Both MANET routing protocols and BGP/OSPF do not have
mechanism to use periodicity of the links to compute paths.

They use only the active links to compute a path

They cannot adopt to network dynamics without requiring a
manual intervention.

Though they are known to be highly stable and scalable, they
can not be directly used in the context of space networks.
Siva kumar Tanguturi & Sanjaya Gajurel
Approaches

Data-Centric approach can be used to enable energy efficient
and low latency operation in proximity networks.

Position Based Routing approach can be efficiently used in
the space orbital and backbone networks having predictable
trajectories.
Siva kumar Tanguturi & Sanjaya Gajurel
Data-Centric & Position-Based
Routing Approach

In Data Centric approach a message specifies its content in
terms of attributes: location, temperature and so on and there
will be a in-networking processing of data.

The Positional Link-trajectory State (PLS) Protocol that is
used can get the link trajectories along with their metrics such
as latency, data rate, error characteristics from STK (Satellite
Tool Kit) to provide the future routing information. Each
node calculates the shortest path and this information is
disseminated throughout the space network.
Siva kumar Tanguturi & Sanjaya Gajurel
Agenda







Introduction to Space Networks
Background
Architecture
Implementation
Simulation Experiments
Conclusions
Discussion
Siva kumar Tanguturi & Sanjaya Gajurel
ASCoT
ASCoT – Autonomous Space Communication Technology
Is a routing and scheduling substrate for flexible
tasking and coordination among space assets.
 Scalable
 Able to deal with message propagation latencies.
 Support connectivity changes
 Support Heterogeneous and asymmetric link
bandwidths
Siva kumar Tanguturi & Sanjaya Gajurel
Assumption
ASCoT expects the underlying system to provide a
variety of information and services to the ASCoT
middleware.
This includes:
 Navigation information



characteristics of the links available and nodes on the other end,
including position (current and expected), bandwidth, reliability,
latency, etc.
Current position of the node
Local status

power, health, load of transmission queues,etc.
Siva kumar Tanguturi & Sanjaya Gajurel
Data-Centric Approach


Naming the data allows the system to eliminate different
levels of binding
Naming the data allows in-network processing of data.
Siva kumar Tanguturi & Sanjaya Gajurel
Data-Centric Approach
Example
Query for the average temperature of the shadowed parts in Gusev crater,

Query can be flooded in the proximity network

only the nodes that meet the query criteria (in Gusev crater and in shadowed
parts) will respond to the query thereby avoiding multiple steps of binding
and spending energy transmitting data from nodes that are not in the
shadowed parts of the crater.

Intermediate nodes also have the context to transform the data in several
interesting Ways


to aggregate different data items that perhaps have redundant information (e.g.
temperature data from nearby sensors)
reduce information in response to resource constraints (e.g. downsample an
image because the image size exceeds available network capacity).
Siva kumar Tanguturi & Sanjaya Gajurel
Data-Centric Approach



Used in Proximity Networks
Energy Efficient
Reduce the latency of communications
Siva kumar Tanguturi & Sanjaya Gajurel
Positioning Link-trajectory State (PLS)

PLS is modified to the context of space networks.

Each node independently computes shortest path tree using
modified Dijkstra’s Algorithm, getting metrics like latency,
data rate, and bit error rate (Satellite Tool Kit,STK).

Unlike traditional Link-State Routing (LSR), the information
disseminated throughout the network is the trajectory of
nodes in space, and the availability of the link end-points now
or in the future.
Siva kumar Tanguturi & Sanjaya Gajurel
PLS

The PLS is only run in mobile space assets like satellites,
moving base stations.

PLS routing exchange information like
{u,p(u),v,t,metrics(u->v)} which are flooded throughout the
network.?

Tradeoffs between frequency of information exchange and
network resources (energy) as well as updates accumulation.

Intelligent scheduling can be employed by which better links
are waited for QoS.
Siva kumar Tanguturi & Sanjaya Gajurel
Key Components of ASCoT

Link Information Dissemination


Path Computation



Takes advantage of the space assets relative predictability by
distributing information about link availability throughout the
network ahead of time.
predicted link connectivity and time-varying graphs are taken into
account for path computation.
intelligent scheduling to meet the application’s QoS requirement
Message Forwarding

once routing table is populated for a given metric, lookup the best
next hop towards the destination and buffer the packet until the link
becomes available.
Siva kumar Tanguturi & Sanjaya Gajurel
Message Switching



The base station acts as a message gateway.
It decodes the data-centric name for the target, encodes it as
an attribute (say temperature) along with other constraints for
the query.
Diffusion Semantics is used to harvest data as follows:



A node translates the query into a interest message, floods the
network & sets up gradients (navigator) in the network.
Nodes (sources) reply the query as attribute-value tuple and inject it
into the network.
Gradients now guide the data to the base station by matching
attributes in the data message to that of the gradients established by
the interest messages.
Siva kumar Tanguturi & Sanjaya Gajurel
Implementation Structure



Web-Based Query
Interface
Implemented in OPNET
Data taken from STK
Implementation in OPNET
Siva kumar Tanguturi & Sanjaya Gajurel
Implementation Structure





ascot_app
ascot_nav
ascot_router
position_manager
Antenna modules
Siva kumar Tanguturi & Sanjaya Gajurel
Demonstrated ASCoT Features

Its ability to deal with heterogeneous hardware. The Earth
and Mars relay satellites, as well as the Mars base station,
may utilize completely different transmission hardware. As
long as they have a form of the IP stack and ASCoT running
on top, the communication occurs seamlessly.

The reliability of the protocol in the face of dynamic network
topology, short link duration and long link latencies. As relay
stations become occluded or occupied with tasks of higher
priority (or orientation requirements force them to cut the
current link), ASCoT automatically selects a different path
that uses orbiters that become available.
Siva kumar Tanguturi & Sanjaya Gajurel
Demonstrated ASCoT Features

PLS routing exploits future link information to predictively
route on paths that become available just as the message
travels along, and buffers messages as it waits for the links to
come up if necessary.

Automatic and efficient path discovery and link information
distribution that allows PLS path computation to occur.
Several parameters allow this behavior to be tuned to the
current
network state.
Siva kumar Tanguturi & Sanjaya Gajurel
Demonstrated ASCoT Features
Source: http://scp.grc.nasa.gov/images/portfolio/an/an_3.jpg
Siva kumar Tanguturi & Sanjaya Gajurel
Simulation Components






PI : Specifies the source and the constraints of the data using
a web interface.
DSN (Deep Space Network): The Madrid DSN node
participated in PLS.
Earth Orbiters: Six Middle Earth Orbit (MEO) satellites
Mars Orbiters: Three satellites in Aerosynchronous and five
satellites in moderately inclined lower Mars orbit.
Mars Base Station: communication with rovers happen
thorough base station
Surface Rovers: “Spirit” and “Opportunity” can
communicate with base station.
Siva kumar Tanguturi & Sanjaya Gajurel
Simulation and Experiments (1)
Siva kumar Tanguturi & Sanjaya Gajurel
Simulation and Experiments (2)
Siva kumar Tanguturi & Sanjaya Gajurel
Simulation and Experiments (3)
Siva kumar Tanguturi & Sanjaya Gajurel
Conclusions




ASCoT is a new data-centric and position-based routing
architecture for future space science mission. The space
missions involve large number of satellites and other nodes
and the current static routing (manual ) is no more scalable.
Data-centric approach avoids the traditional address-centric
energy consuming approach to make up for the energy
deprived space nodes.
Planning to add design scheduling and resource allocation
strategies.
Even with the limited knowledge about the future available
links, ASCoT can discover paths that can be used to forward
message successfully and efficiently.
Siva kumar Tanguturi & Sanjaya Gajurel
Critiques




This paper tries to solve the communication difficulties in
space network by emphasizing the data-centric and positionbased routing approach.
The data to be communicated between the earth and the Mars
is only the telemetry type. Also didn’t address the issues of
real time and bulk load (picture, video) data transfers.
Direct communication facilities among the surface elements
required for the space mission has not been mentioned.
Using PLS, the router is queuing packet when the satellites
get occluded but didn’t mention how long. That can be hours
and special store and forward router (used in DTN) may be
required.
Siva kumar Tanguturi & Sanjaya Gajurel
Question Session


Feel free to ask the doubts and questions. We
will try to answer them 
Your comments are really appreciated
Thank You
Siva kumar Tanguturi & Sanjaya Gajurel