Transcript 22 September 2008 2

Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Technical Kick off meeting
Grosuplje September 22-23 2008
System Architecture
Elwyn Davies
[email protected]
www.folly.org.uk
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Agenda
•
•
•
•
•
•
•
•
Terminology
N4C Architectural Scenario
Node Type Taxonomy
Application Taxonomy and Issues
Node System Partition
Sub-systems Beyond the Node
Review Criteria for WP2
Architectural Issues
22 September 2008
www.n4c.eu
2
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Terminology
• Round Trip Bound (RTB)
– Maximum time for a network round trip
• Legacy Internet (LI)
– The 'core' connected part of today's Internet
– Environment with low elasticity on RTB
• Communication Challenged Realm (CCR)
– Area that needs DTN or some other technique
– Environment with high elasticity on RTB
• Enclave
– Area with CCR that can use LI techniques locally but is not
connected to LI core with low latency links
22 September 2008
www.n4c.eu
3
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
N4C
Scenario
Legacy
Enclave
Mobility of nodes
enclave <-> DTN
DTN region 2
Coordination of
gateways
Mobility of nodes
legacy<-> DTN
DTN region 1
Communication
within DTN
DTN-DTN
communication
Key
Legacy Internet
Region with 'traditional' connectivity
Region with only DTN connectivity
Enclave in DTN region
Gateway
DTN/Legacy Node in DTN
22 September 2008
www.n4c.eu
Communication
between DTN
and Legacy
DTN/Legacy Node in Legacy
Legacy only Node
Communication path
Node mobility excursion
Coordination functionality
4
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Kinds of Node
(aka Node Taxononomy)
• Pure Legacy Nodes
– 'Standard' Internet machine maybe with adapted applications
running in 'compatibility mode'
• DTN Only Nodes
– Sensor group leaders and static DTN 'routers'
• Chameleon Nodes
– mobile, capable of running in both CCRs and LI
• Gateway Nodes
– Generally static nodes linking CCR and LI
• Ability of DTN nodes to provide bundle 'custody'
– May be a significant extra categorization
22 September 2008
www.n4c.eu
5
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Application Taxonomy
• Store and Forward Paradigm
– e.g., email, maybe peer-to-peer applications
• Client-Server Paradigm
– Request-response mechanism, e.g., WWW
• Unidirectional Real Time Stream Paradigm
– e.g., time-shifted TV programme
• Interactive Real Time Stream Paradigm
– e.g.. Skype
22 September 2008
www.n4c.eu
6
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Application Issues
• Many current applications rely on running in
low elasticity RTB environment (i.e., LI)
• Adapting applications requires
– Technical solution for DTN working
– User Expectation Management (different UI?)
• Some applications are never going to work
– especially Interactive Real Time Streams
22 September 2008
www.n4c.eu
7
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Node Sub-system Partition
Management and Control Interface
Application Interface
Security
Support
Bundle
Storage
Gateway
Support
Chameleon
Management
Module
Manager
Transport Select
Alternative
Transports
LTP
Storage
Management
DTN Bundle
Protocol
Comms
Opportunity
Management
Network Multiplex
DTN
Routing
Address
Cache
Network Interface (Wireless Interface(s))
22 September 2008
www.n4c.eu
8
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Gateway Coordination
• Address mapping
• Bundle to packet & packet to bundle conversion
• Inter-gateway coordination to manage arrival of the
same data at multiple gateways (especially on the
CCR side)
– Needs a protocol.
• Redirection of sessions/bundles/packets that need to
be injected into a different CCR through another
gateway or in to the same CCR via a more
appropriate gateway.
22 September 2008
www.n4c.eu
9
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Address Management
• Nodes in CCRs don't have benefit of DNS
– need more permanent cache of mappings
• Cache management
– DTNmail has rudimentary cache
• BUT update scheme is not robust
• Identifier/Locator Issue (discussion point!)
– Unique Node Identfier
– Multiple locators (for Chameleon Nodes)
• One for LI
• One for each CCR where it is a 'member'
22 September 2008
www.n4c.eu
10
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Routing:
within CCR + to/from LI
• Separate session on Routing...
– so just an outline
• PRoPHET
– implementation improvements and tuning
– possibly extensions for
(loosely) scheduled 'opportunities'
• Experiments with alternative schemes?
• Integration with Gateway Coordination
22 September 2008
www.n4c.eu
11
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Node Management
• Must be able to manage nodes remotely and locally
• Must be able to monitor nodes (ditto)
• Must be able to install updated modules over the
DTN
• Must be able to install alternative functionality in
place of current modules
• Must not overload the DTN network with logging
22 September 2008
www.n4c.eu
12
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Review Criteria
Infrastructure must provide...
•
•
•
•
Forwarding of DTN bundles as specified in [RFC5050]
Support for DTN transport protocols, including at least LTP, between adjacent nodes
Routing mechanisms suitable for directing DTN bundles through the CCR
Gateway mechanisms
– to mediate inter-realm communications between the core LI and CCR DTN regions
– to mediate inter-realm communications between CCR DTN regions and CCR enclaves using LI
protocols
– manage linkages between realms where multiple gateways are provisioned.
•
•
•
•
•
•
Manage configuration and addressing of nodes providing DTN functionality, and the
mapping of addresses between the LI and DTN forms where necessary
Manage mobility of chameleon nodes moving from LI realms to DTN realms and vice
versa
Support for authentication of nodes and messages in the DTN environment
Support for encryption of messages passing through the DTN environment
Logging and monitoring of the infrastructure
Mechanisms for secure distribution and installation of new components over the DTN
infrastructure
22 September 2008
www.n4c.eu
13
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Architectural Issues
•
•
•
•
•
What is the nature of a CCR?
Seamless Mobility
Transport Integration
Addressing Schemes, Mappings & Caching
Clock Synchronization
22 September 2008
www.n4c.eu
14
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
The Nature of an (N4C) CCR
• Topology vs Geography
– LI primarily defined by topology
(stretched by wireless)
• "Do I have an IP(v4) address?" (membership token)
• "Can I get a connection?"
• But what about CCRs?
– "Doh! It's a (geographical) region, stupid!"*
– OK, but: what defines membership? ....
• What if members meet outside the 'region'?
• How many CCRs are there?
* D.T.N. Simpson, 2002
22 September 2008
www.n4c.eu
15
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Addressing
Identifiers and Locators?
• Nodes need a unique identifier
– BUT are multiple 'locators' needed to 'find' a node either in
the LI or one of many CCRs?
• does this tie up with CCR 'membership'?
• is it easier to control delivery predictabilities in a CCR based on a
separate locator space?
• is it easier to control replication/looping with separate locators?
• How to manage robust distribution and caching of
identifiers, locator mappings and associations with
particular CCRs in a DTN environment?
– and to manage the linkage of users to nodes?
22 September 2008
www.n4c.eu
16
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Seamless Mobility - 1
• A User Perception....
– implies requirements on applications
• need to manage user expectations
• Coping with different transport paradigms
– Alternatives (may be best to offer both!):
• Hide the differences in transport layer
Is this desirable? ... possible always?
• Let the application know and handle the differences
but aim for maximum commonality of interface
22 September 2008
www.n4c.eu
17
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Seamless Mobility - 2
• How does a node/application know 'where' it is?
– Differentiating between Wi-Fi in DTN and LI?
– Does node need to know its 'current' CCR?
– How does it 'join' a new CCR?
• How to handle connectivity from LI with a node in a
CCR?
– Gateways provide an 'anchor' in the LI
– Comparisons with
• SIP registration/Session Border Controllers (arrggggh!)
• Mobile IP(v6) home agents
22 September 2008
www.n4c.eu
18
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Transport Integration
• Can/Should Bundle Protocol be integrated into
transport layer?
• How to offer an (extended) socket interface
for all DTN transports?
• Would it be better to use LTP for bundle
exchange at opportunistic encounters?
– Might imply LTPBundleLTP layering!
22 September 2008
www.n4c.eu
19
Networking for Communications Challenged Communities:
Architecture, Test Beds and Innovative Alliances
Grant Agreement: 223994
Clock Synchronization
• Bundle Protocol relies on reasonably accurate
wall clock time in all nodes
• This may be a problem for nodes that need to
run for long periods with little external contact
• Can requirement be removed?
22 September 2008
www.n4c.eu
20