PPT file of slides

Download Report

Transcript PPT file of slides

Content Centric Networking in
Tactical and Emergency MANETs
Soon Y. Oh, Davide Lau, and Mario Gerla
Computer Science Department
University of California, Los Angeles
{soonoh, chiume, gerla}@cs.ucla.edu
Introduction
 Infrastructureless nature and quick deployment  a
MANET is ideally suited for emergency & tactical
operation, but
 Challenging environments
 Lossy channel and high mobility
 Limited resources
 Hard to find necessary content
 No search engine
 Scalable & efficient content search and dissemination
in MANETs  Content Centric Networking
2
Content Centric Networking (CCN)
 Users are interested in WHAT content – not
WHERE it is or WHO has it
 Data is addressed by NAME OR CONTENT –
rather than by location or IP address
 No overhead in binding name to location
 Enabled by low storage prices and high speed links
Can CCN be directly applied to
MANET environment?
3
WiCCN = CCN in MANETs
 Advantages
 Group based mobility/operation
 resource sharing within group
 Hierarchical data structure
 Information locality (via Cache)
Content Centric
Networking
 Challenges
 Lossy channel and resource shortage
 Data Push and Pull is required while Internet CCN is only Pull
 Must Push Critical information and operation messages
 Security and content authentication
 Critical data and wireless broadcast medium
4
WiCCN protocol design goals
 Hierarchical storage/search architecture
 Topic based data vs spatial/temporal contents
 Cross-layer approach
 Scalable and resource aware
5
Related Work
 TRIAD (2000)
 User-friendly, structured, with location-independent names and content
addressing (has influenced later protocols)
 Data-Oriented (and beyond) Network Architecture (DONA) (2007)
 Flat, self-certifying names instead of IP addresses and DNS
 Contents is published and registered with a tree of trusted Resolution
Handlers (RH)
 Routing on Flat Levels (ROFL) (2006)
 Semantic-free flat labels; it creates a circular namespace, e.g., DHT
 Content Centric Network (CCN) (2009)
 Network wide content caching and user-friendly, hierarchical names for
routing; Digital signature for security
 Named Data Networks (NDN) (2010)
 Future Internet Architecture
6
WiCCN Network Model




Group based mobility
Hierarchical topology
Interconnection via gateways
Heterogeneous devices – different capacities
Airborne
Network
Wideband
Network
Soldier
System
7
WiCCN Content Types
 Topic based content
 Data files, video and audio clips
 Data is stored at publisher (originator) or near backbone nodes and
travels anywhere in the network
 PULLED by users
 No location and time sensitivity
 Spatial/temporal content
 Situation awareness data; operational messages
 Content value is time and location sensitive
 PUSHED by publisher towards command center or proper
location
8
Local Storage
Content
Repository
 Content Repository
 Intermediate nodes cache content
 Maximize the probability of sharing
Meta-Data Registry
 Meta-Data Registry
 Hash table for efficient look up
Interest Table
 It is used to forward Interest packet
 Meta-Data includes content attributes, e.g., type, time, loc, etc
 Interest Table
 Stores Interest Query packets
 To suppress duplicate Interest packets
 To relay content to requestors
9
WiCCN Routing
 Content Pushing
 Spatial/temporal content
 Geo-routing to command center or other destination
10
WiCCN Routing (Cont.)
 Content Pulling
 Using an Interest packet and local storages
1. Check Content Repository and send data
if it exists
2. If there is no content, check Meta-Data
Repository
3. If Meta-Data entry exist, a node relays
Interest toward data origin
4. Otherwise, Interest is passed to a
Gateway toward upper level
Content
Repository
Interest
Meta-Data Registry
Interest
5. Interest is relayed
Interest Table
11
WiCCN Routing (Cont.)
 Difference to Internet CCN (due to wireless
common medium)
 Interest aggregation
 Time stagger re-broadcast Interest packets
 Upon overhearing the same Interest, cancel the re-broadcast
 Data Packet collision avoidance
 If more than one neighbors tries to transmit
 Exchange Request/Reply
 Respond with Reply before transmitting data
12
Packet Collision Avoidance
REPLY
Content
Interest
REQUEST
REPLY
13
Security and Authentication
 Using PKI
 A gateway has private key and members in the
domain have public keys
 A gateway adds digital signature using a private
key
 Members encrypt packets using the public key
 The private and public keys are pre-assigned
14
Implementation
 Implement WiCCN on Linux OS
 A gateway and members
 The gateway floods/updates meta-data
 A node sends Interest
 Request/Reply- exchange and data transmission
 Run simple four node topology
 Compare performance with peer-to-peer protocol,
e.g., Pastry over OLSR
15
Pastry Overhead




Every 3s new data generated (no real data transmitted)
A gateway floods meta-data
Pastry 378B/s average overhead
Traffic suddenly increases to maintain a P2P ring structure
 OLSR traffic in the background
16
WiCCN Overhead




Every 3s new data generated (no data transmission in this experiment)
A gateway floods meta-data
Pastry 72B/s average overhead
Only Meta-Data flooding
17
End-to-End Delay
 From node A to node D in the 4 node chain topology
 File size 1, 5, 10, 15, 20, 25, 100MB
 Pastry and WiCCN experience same delay in peer to peer transmissions
18
End-to-End Delay (Cont.)




From node A to all nodes in the previous 4 node topology
No broadcast; each node requests data at different time
WiCCN presents significant lower delay due to content caching
In Pastry, node A transmits 3 times, but WiCCN node A transmits only once;
cached data, at an intermediate node, is transmitted
19
Conclusion
 WiCCN performs better than DHT based
content sharing
 Mainly due to caching
 Future work:




Implement on smart phones
Experiment with mobility
Design cache strategies
Bigger testbed/emulator
20
21