implementation of ip mesh network using ule protocol over dvb-s links wan tat chee network research...
TRANSCRIPT
Implementation of IP Implementation of IP Mesh Network using ULE Mesh Network using ULE
Protocol over DVB-S Protocol over DVB-S LinksLinksWan Tat CheeWan Tat Chee
Network Research Group Network Research Group School Of Computer SciencesSchool Of Computer Sciences
Universiti Sains MalaysiaUniversiti Sains Malaysia 11800 USM, Penang, Malaysia.11800 USM, Penang, Malaysia.
[email protected]@cs.usm.my
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 22
IntroductionIntroduction
Satellite-based Internet connectivitySatellite-based Internet connectivity• Provides access to rural / underserved Provides access to rural / underserved
communitiescommunities• Able to operate in areas without existing Able to operate in areas without existing
telecommunications infrastructuretelecommunications infrastructure• Quick setup for disaster relief and Quick setup for disaster relief and
recovery efforts (e.g., earthquakes, recovery efforts (e.g., earthquakes, tsunami, etc.)tsunami, etc.)
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 33
IssuesIssues Existing satellite-link based Internet access Existing satellite-link based Internet access
equipment is expensiveequipment is expensive Existing technology has inefficient spectrum Existing technology has inefficient spectrum
usageusage Existing hub-based network averages two hops Existing hub-based network averages two hops
among nodes (high latency)among nodes (high latency) DVB-S is a cost-effective method used in satellite DVB-S is a cost-effective method used in satellite
TV broadcast (e.g., Astro)TV broadcast (e.g., Astro) However, It is unidirectional in natureHowever, It is unidirectional in nature Unidirectional Link Mesh Network implements Unidirectional Link Mesh Network implements
bidirectional network access using DVB-S bidirectional network access using DVB-S technologytechnology
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 44
Existing Point-to-Point Existing Point-to-Point Satellite LinksSatellite Links
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 55
UniDirectional Link Mesh UniDirectional Link Mesh Networking ApproachNetworking Approach
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 66
Hardware RequirementsHardware Requirements
An ASI (Asynchronous Serial Interface) An ASI (Asynchronous Serial Interface) card to transmit MPEG2-TS packet (Linear card to transmit MPEG2-TS packet (Linear Systems’ DVB Master FD).Systems’ DVB Master FD).
(N – 1) DVB receiver cards where N is the (N – 1) DVB receiver cards where N is the number of nodes in the UDL mesh number of nodes in the UDL mesh network. List of supported DVB receiver network. List of supported DVB receiver cards: cards: http://www.linuxtv.org/wiki/index.php/DVB-http://www.linuxtv.org/wiki/index.php/DVB-S_cardsS_cards..
A DVB-S modulator (Newtec NTC2177).A DVB-S modulator (Newtec NTC2177).
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 77
Node Configuration Node Configuration (for 3-Node WAN network)(for 3-Node WAN network)
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 88
Configuration for 3 Nodes Test BedConfiguration for 3 Nodes Test Bed
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 99
Bidirectional DVB-S Bidirectional DVB-S ULE ArchitectureULE Architecture
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 1111
Experiment and Result (cont)Experiment and Result (cont)Throughput for TCP Unicast v.s. TCP Window Size
0
200
400
600
800
1000
1200
1400
1600
1800
2000
1 14 27 40 53 66 79 92 105 118 131 144 157 170 183 196
Time (s)
Th
rou
gh
pu
t (k
bp
s)
TCP WS216KBytes
TCP WS180KBytes
Log. (TCPWS216KBytes)
Log. (TCPWS180KBytes)
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 1212
Experiment and Result (cont)Experiment and Result (cont)
Throughput Comparison for TCP Unicast on 8Mbps DVB-Mesh v.s. Benchmark on 10Mbps Wired LAN
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
1 15 29 43 57 71 85 99 113 127 141 155 169 183 197
Time (s)
Th
rou
gh
pu
t(kb
ps)
Benchmark10Mbps WiredLAN TCP WS216KBytes
8Mbps UDL-Mesh TCP WS216KBytes
Log.(Benchmark10Mbps WiredLAN TCP WS216KBytes)Log. (8MbpsUDL-Mesh TCPWS216KBytes)
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 1313
Experiment and Result (cont)Experiment and Result (cont)Throughput for UDP Unicast v.s. Offered Load
0
1000
2000
3000
4000
5000
6000
7000
1 2 3 4 5 6 7 8 9
Time (s)
Th
rou
gh
pu
t (k
bp
s)
1Mbps
2Mbps
3Mbps
4Mbps
5Mbps
6Mbps
7Mbps
8Mbps
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 1414
Experiment and Result (cont)Experiment and Result (cont)
Average Packet Loss for UDP Unicast
1Mbps0%
2Mbps0%
3Mbps0%
4Mbps0.18%
5Mbps16%
6Mbps29%
7Mbps32%
8Mbps40%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Bandwidth Type
Pac
ket L
oss
(%)
1Mbps
2Mbps
3Mbps
4Mbps
5Mbps
6Mbps
7Mbps
8Mbps
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 1515
Experiment and Result (cont)Experiment and Result (cont)
Throughput Comparison for TCP Unicast and UDP Unicast
0
500
1000
1500
2000
2500
3000
3500
4000
4500
1 13 25 37 49 61 73 85 97 109 121 133 145 157 169 181 193
Time (s)
Th
rou
gh
pu
t(k
bp
s)
TCP (WindowSize216KBytes)
UDP(Bandwidth4Mbps)
Log. (UDP(Bandwidth4Mbps))
Log. (TCP(Window Size216KBytes))
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 1616
Analysis of ResultAnalysis of Result
All UDL nodes are allocated 8Mbps of All UDL nodes are allocated 8Mbps of bandwidth for the tests.bandwidth for the tests.
The throughputs of TCP and UDP traffic are The throughputs of TCP and UDP traffic are about 1/2 and 1/8 of allocated bandwidth about 1/2 and 1/8 of allocated bandwidth respectively.respectively.
ASI card buffer queue affects the latency ASI card buffer queue affects the latency of packet sending. Less buffer queue of packet sending. Less buffer queue decreases latency, but increases CPU decreases latency, but increases CPU usage due to more interrupts and context usage due to more interrupts and context switches.switches.
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 1717
Analysis of Result (cont)Analysis of Result (cont)
Faulty ASI card firmware disables Faulty ASI card firmware disables automatic null packets insertion. Thus automatic null packets insertion. Thus bidirectional module need to insert null bidirectional module need to insert null packets manually.packets manually.
Consequently, real data packets have to Consequently, real data packets have to compete with null packets for resources. compete with null packets for resources. Thus the performance is less than ideal.Thus the performance is less than ideal.
Due to high latency, bigger TCP window Due to high latency, bigger TCP window size results in better throughput.size results in better throughput.
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 1818
Future WorkFuture Work
Reimplemention of bidirectional Reimplemention of bidirectional module as a kernel space module to module as a kernel space module to cut the cost of context switches and cut the cost of context switches and improve performance.improve performance.
Try to obtain an ASI card that Try to obtain an ASI card that properly insert null packets properly insert null packets automatically.automatically.
April 12 April 12 (c) 2006 TC Wan(c) 2006 TC Wan 1919
Links & ContactsLinks & Contacts
IP over DVB-S via ULE research page: IP over DVB-S via ULE research page: http://nrg.cs.usm.my/satellite_ule.htmhttp://nrg.cs.usm.my/satellite_ule.htm
Bidirectional module research page: Bidirectional module research page: http://nrg.cs.usm.my/ule.htmhttp://nrg.cs.usm.my/ule.htm
Coordinator: Coordinator: [email protected]@cs.usm.my Developers: Developers: [email protected]@nrg.cs.usm.my, ,
[email protected]@nrg.cs.usm.my, , [email protected]@nrg.cs.usm.my..