Download presentation
Presentation is loading. Please wait.
1
ECE 544 Protocol Design Project 2016
Wuyang Zhang Yu Sun Shuo Fan
2
Network Architecture & Topology Assumptions
Service Objective: k-out-of-n “packet datagram” multicast
3
Network Architecture & Topology Assumptions
Unreliable network with packet loss prob. per link of p Each end node is attached to only one router All links have same characteristics (i.e. hop cost 1 and same MTU 1500 bytes) Small maximum number of nodes (i.e. 50) …others – Network is stable and without mobile devices. Network address is static. No transmission failure. 3
4
Protocol Concept Overall concept of your proposed protocol
Multicast message to “best k (=<n)” destinations with the minimal total hop count Summary of protocol design (packet structure, control plane for protocol, data plane for forwarding, key algorithms for mcast) ARQ Scheme After considering about the performance and implementaion, we finally choose Go back N ARQ 4
5
Syntax – Packet Formats
Control Packets & formats Control messages Hello message (send link probes to neighbors) ACK message (response to neighbors) routing topology UPDATE Data packet format and headers SRC ADDR multicast DEST ADDR1, …...multicast DEST ADDR k Number of candidate destination address Number of target receivers ( the value of K) Length of payload
6
Semantics Hello & ACK Message Control Type: HELLO Type = 1
ACK Type = 2 UPDATE Type = 3 8 16 24 31 Control Type Device Type SRC ADDR DEST ADDR
7
Semantics Update Message
Update message: Src Addr already knew N paths and want to tell these topology to its neighbor(Dest Addr) Path1 = x: Src Addr knows one path, which consists of x nodes, and Src Addr knows each Addr during every hop count. Update Message 8 16 24 31 Control Type Update Sequence Src Addr Dest Addr Path1 = x Addr1 Addr2 Addr3 Addr4 … Addr x Type x Path 2 = y ... Path N = m Addr m Type m
8
Semantics Data Packet Format
8 16 24 31 Control Type Src Addr NCDA NTR Dest Addr 1 Dest Addr 2 Dest Addr N Length PAYLOAD NCDA: Number of Candidate Destination Address (initially = N) NTR: Number of Target Receivers = K
9
Example Networks Network Topology
By the given network topology ,we show how to build up the routing table and the forwarding table, source-specific tree and K out of N algorithm. R5 R6 How to build up the routing table, forwarding table?
10
Routing Table R1 R2 R3 R4 R5 R6 / 1 R3-R4 R2-R4 R2-R4-R3 R3,R5 R2-R4-R3-R5 R4-R3 R1-R3 R1-R3-R5 R4-R3-R5 R4-R2 R1-R2 R3-R1 R3-R5 R5-R3 R5-R3-R1 R5-R3-R4 going to build up the routing table. -> Each router can know how to reach to other routers and hosts from the update message. We fetch the routing information from the update messages. But unlike the distance vector, instead of only recording the next hop to destinations, we write the complete routing paths in the routing table.
11
Forwarding Table R1 Source Dest Path Cost Interface S D1 D2,D3 R2-R4 R2-R4-R3-R5-R6 2 5 Interface 1 R3-R4 R3-R5-R6 3 Interface 2 R2 Source Dest Path Cost Interface S D1 D2,D3 R4 R4-R3-R5-R6 1 4 Interface 1 R1-R3--R4 R1-R3-R5-R6 3 Interface 2 After building up the routing table, let’s see how to construct the forwarding table based on the routing table. Since one of our assumption is the destination is connected by only one router, the problem can be transferred to how the source point find the forwarding path to that access router. From the routing paths recorded in the routing tables each router can find the complete forwarding path to R4 and R6. Also since each link has the same cost, the total cost of the source point to get the destination is the hop count which is the number of routers connecting two of them. R5 Source Dest Path Cost Interface S D1 D2,D3 R3-R4 R6 2 1 Interface 1 / Interface 2 11
12
Source-Specific Tree Motivation: For every specific source, we build a tree from the network topology that allows the least total hop count from the source to all destinations. ( Satisfy the situation when K equals to N) Basic Principle: The source specific tree should cover the most shared paths to destinations. After having the forwarding table for each router to get the destinations, the next step is to construct the source-specific tree based on the network topology. Algorithm of building the source-specific tree: 1.Count the number of shared paths to all destinations among possible path pairs 2.Subtract shared hops from unicast hops to all destinations and get the total hop count for a certain path combination ( total unicast hops - shared hops) 3.Compare all path pairs, select the path with the minimal final hop computed in step 2
13
Select K out of N ( 4 scenarios )
Scenario 1: no path is shared among N destinations R1 Action: Select the K nearest destinations
14
Getting K out of N ( 4 scenarios )
Scenario 2: All N destinations share a path Action: Unicast to the next router
15
Getting K out of N ( 4 scenarios )
Scenario 3: 2 out of 3 destinations share paths. algorithm complexity is high, consider the number of path pairs. Action: Compare <total hop count to D2,D3>, < hop count to D1 + min hop count(D2, D3)>
16
Getting K out of N ( 4 scenarios )
Scenario 4: 2 out of 3 destinations share paths. D1 D2 R1 D3 Action: Compare <total hop count to D2,D3>, < hop count to D1 + min hop count(D2, D3)>
17
Summary k-out-of-n “packet datagram” multicast
Key protocol design features k-out-of-n “packet datagram” multicast Performance Low throught because of only 1 byte address length Each SRC ADDR tracks every known path in detail We could finally find the optimal shortest path based upon searching shared path Implementation complexity There will be at most N different paths from source to destination, and the number of destination host is 3 so time complexity is less than O(N^3) 17
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.