Routing, Resource Discovery and Hierarchical Architecture in Wireless Networks
description
Transcript of Routing, Resource Discovery and Hierarchical Architecture in Wireless Networks
Routing, Resource Discovery and Hierarchical Architecture in Wireless Networks
Ahmed Helmy
Lecture 4 - A
Spring 08
Birds-Eye View: Research in Wireless Networks @ UFL
Architecture & Protocol Design Methodology & Tools
Test Synthesis (STRESS)
Mobility Modeling(IMPORTANT)
Protocol Block Analysis (BRICS)
Query Resolution in Wireless Networks (ACQUIRE & Contacts)
Robust Geographic Wireless Services (Geo-Routing, Geocast, Rendezvous)
Gradient Routing (RUGGED)
Worms, Traceback in Mobile Networks
Behavioral Analysis in Wireless Networks
(MobiLib & IMPACT)
Multicast-based Mobility (M&M)
Mobility-Assisted Protocols (MAID)
Context-Aware Networks
Road Map
• Unicast Routing with caching
• Resource Discovery issues
• Contact based architectures– CARD, TRANSFER
• Geographic routing issues
Related Reading list• From book (C. Perkins): (one order of reading)
– ch 3 DSDV, ch 5 DSR, ch 6 AODV, ch 8 TORA, ch 4 cluster-based, ch 7 ZRP
• From the paper reading lists– unicast routing list in syllabus (all papers, most overlap with
book, read the book or papers)– broadcast: - broadcsat storm, - min dominating sets– Resource discovery:
• book chapter (on-line)• TRANSFER, CARD, MARQ
Related Reading list (contd.)– Mobility Modeling
• IMPORTANT, PATHS, BRICS, Bk Chptr (posted)
• Partial reading list for geographic routing aspects– GPSR, - LAR, - Geocast, – From my web site
• Perfect Geocast
• Inaccuracy/inconsistency ,
• Face routing with inaccuracy
• Mobility prediction
• Black listing
• Book chapter (posted)
Routing and Resource Discovery in Wireless Networks
Conventional Approaches
• Flooding:– Simple– Overhead: N-1 transmissions, g.(N-1) receptions
(where g is average node degree)… expensive!
• Expanding Ring Search:– Repeated floods with expanding TTL– Terminates when target found– Initial TTL & increment have significant impact
Conventional Approaches (contd.)
• Reduced Broadcast heuristics (bcast storm ppr):• Probabilistic rebroadcast with p
• Counter-based: suppress bcast if cnt rx > cnt thresh
• Distance-based: suppress if min(d) < D thresh, where min(d) is min dist to nbr bcast
• Location-based: supp if added coverage > Area thresh
– Attempts delivery to all nodes– Does not guarantee delivery– Works well when density/redundant bcast is high
• Cluster-based and Min. Dominating Set
- Guarantees delivery - Optimum Dist Algo NP hard- Approximations reduce amount of redundant bcasts to cover all the nodes- re-configuration may be major with dynamics/mobility
Example connected dominating set, black nodes belong to the
backbone constituting a dominating set.
• On-demand routing with caching
1
2
3
Q
T
T
4.b
4.a
Cache table
T: target nodeQ: queriertr: transmission range request reply
tr2
2
2
3
3
3
4.b
• DSR-like routing model:
– a querier node Q issues a request for a target resource T.
– The request process progresses as follows:
• 1. Q performs local lookup,
• 2. If a cached route to T is not found then request a lookup from 1-hop neighbors (within transmission range),
• 3. If a cached route is not found (or is invalid) then flood a request throughout the network,
• 4.a. Intermediate nodes with cached route to T reply to Q,
• 4.b. The target T replies to requests returning multiple paths to Q.
Localcachelookup
Flood request
Valid route
Neighborcachelookup
Foundlocalcache
Foundneighborcache
p1 q1
1-p1
1-q1
p2
1-p2 q2 1-q2
qr1- qr
DiscoveryRequest
State diagram for on-demand routing in DSR-like mechanisms.
p:hit ratioq: validity ratioqr: obtained by flooding
Overhead (number of transmitted messages)1 - local cache lookup: no overhead2 - Nbr Cache lookup: NC=1+.g, where g av deg, and fraction of nbrs responding3 - Cost of Flood: CF~(N-1)+g.L+.N..L, where L is av. path length from Q to T, .N is fraction of intermediate
nodes responding, and .L is av. path length for such responses
Ov=(1-p1).NC+(1-(p1q1+(1-p1)p2q2)).CF/qr
Define cache efficacy as p.q (hit a valid local or nbr cache)[Recall: Overhead p.q ]
For small transfers, in mobile, large-scale, wireless networks,cache efficacy is very low, - need new paradigms: - hierarchy, - hybrid/loose hier. Arch.s