on the impact of intrinsic delay variation sources on ...€¦ · i leo orbit, circular and nearly...
TRANSCRIPT
On the impact of intrinsic delay variation sourceson Iridium LEO constellation
Amal BOUBAKER,IRIT/TeSAEmmanuel CHAPUT, INPT Andre-Luc BEYLOT, INPT
Nicolas KUHN, CNES Jean-Baptiste DUPE, CNESRenaud SALLANTIN, TAS Cedric BAUDOIN, TAS
4th of March 2020
Delay variation on Iridium 04/03/2020 1/40
Table of Contents
Context
Different intrinsic sources of delay variation
Simulations and ResultsNS-2 : one flowNS-2 : Multi-flow fairnessOpenSAND
Delay variation on Iridium 04/03/2020 2/40
Table of Contents
Context
Different intrinsic sources of delay variation
Simulations and Results
Delay variation on Iridium 04/03/2020 3/40
Context (1/3)
I What are satellite constellations ?I Satellite constellations : set of satellites communicating (or
not) with each other.I and evolving in several adjacent planes in their orbit
I According to its altitude: VLEO, LEO, MEO ou GEOI or type of the orbit: Elliptic, circular, polar, nearly polar etc ..I or the use: Satellite positioning system, Telecommunication,
Remote sensing etc ..
Figure: Example of a Satellite Constellation : IridiumDelay variation on Iridium 04/03/2020 4/40
Context (2/3)
I Why study these constellations ?I Several existing (rival) MEO and LEO constellations or future
projects.I Need for broadband (or even very high) where terrestrial
solutions have failed ..I Little information on the impact, of the delay variation due to
intrinsic sources in these constellations, on the performance ofthe protocol layers with recent stacks.
Delay variation on Iridium 04/03/2020 5/40
Context (2/3)
Constellation selection criteria :
I LEO orbit, circular and nearly polar.
I 6 orbital planes and 11 satellites per plane.
I With 4 ISL1(therefore satellites communicating with one another)
I and already deployed
90
60
30
0
-30
-60
-900 30 60 90 120 150 180-30-60-90-120-150-180
unprojected longitude
65
64
63
62
61
60
59
58
57
56
55
54
53
52
51
50
49
48
47
46
45
44
43
42
41
40
39
38
37
36
35
3432
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
15
4
3
2
1
0
unpro
jecte
d latitu
de
(a) A trace of Iridium’s topology
DescendingSatelllites
AscendingSatelllites
OrbitalSeam
N
(b) Orbital seam
1Inter-satellite LinksDelay variation on Iridium 04/03/2020 6/40
Table of Contents
Context
Different intrinsic sources of delay variation
Simulations and Results
Delay variation on Iridium 04/03/2020 7/40
Different intrinsic sources of delay variation (1/4)
LEOs are known for short propagation delays compared to MEO orGEO, however they are subject to higher delay variations (jitter)due to the movement of satellites with respect to other satellitesand terrestrial terminals.Delay variations are due to:
1. Satellite elevation:I Variation of the elevation angle due to the movement of the
sat. with respect to the ground terminal.I Frequency: During ∼ 10min the delay varies within [2.6 ms;
8.2 ms].
1
2
43
Ascending satellite
Descending satellite
Transmission not-possible
Delay variation on Iridium 04/03/2020 8/40
Different intrinsic sources of delay variation (2/4)
2. Intra-orbital handover:I When the satellite drops below the elevation mask of the
terminal,(moving or not), the connection is consequentlyhanded over to the following satellite in the same plane (ornot)
I Frequency: Every ∼ 10 min.
1
2
43
Ascending satellite
Descending satellite
Transmission not-possible
3. Inter-orbital handover:I The rotation of earth on its axis or the movement of the
ground terminal along the longitude, give rise to a connectionhandover to another sat. in the adjacent orbital plane.
I Frequency: Every ∼ 2 hours.
Delay variation on Iridium 04/03/2020 9/40
Different intrinsic sources of delay variation (3/4)
4. Seam handover:I A particularity of the polar or nearly polar satellite
constellations is that the satellites in the last and first orbitalplanes do not have any links with each one another.
I Frequency: At most 3 times and at least twice a day, theduration depends on the longitudinal separation of the twocommunicating ends.
1
2
43
Ascending satellite
Descending satellite
Transmission not-possible
5. ISLs changes:I For polar or nearly polar orbits the ISLs are deactivated at the
pole because of the sats. which pass at very high speeds andin opposite directions.
I Frequency: Every ∼ 10 min. And as seen in the Fig.2a 4 sats./ plane are on the poles’ edges.
Delay variation on Iridium 04/03/2020 10/40
Different intrinsic sources of delay variation (3/4)
90
60
30
0
-30
-60
-900 30 60 90 120 150 180-30-60-90-120-150-180
unprojected longitude
65
64
63
62
61
60
59
58
57
56
55
54
53
52
51
50
49
48
47
46
45
44
43
42
41
40
39
38
37
36
35
3432
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
15
4
3
2
1
0
unpr
ojec
ted
latitu
de
0
20
40
60
80
100
120
140
160
0 2 4 6 8 10 12 14 16 18 20 22 240
20
40
60
80
100
120
140
160
Del
ay (m
s)
Time (hr)
4 2
1
3 0
10
20
30
40
50
16 16.2 16.4 16.6 16.8 170
10
20
30
40
50
Dela
y (m
s)
Time (hr)
Figure: The evolution of the OWD2in Iridium and the factors of delay
2One-Way DelayDelay variation on Iridium 04/03/2020 11/40
Table of Contents
Context
Different intrinsic sources of delay variation
Simulations and ResultsNS-2 : one flowNS-2 : Multi-flow fairnessOpenSAND
Delay variation on Iridium 04/03/2020 12/40
Table of Contents
Context
Different intrinsic sources of delay variation
Simulations and ResultsNS-2 : one flowNS-2 : Multi-flow fairnessOpenSAND
Delay variation on Iridium 04/03/2020 13/40
NS-2 : one flow (1/12)
NS-23settings:I Iridium:
I Plane : 6.I Sat. per plane : 11.I Up/Downlink bandwidth : 1.5 Mbps.I ISL bandwidth : 25 Mbps.I Buffer size : BDP4
I TCP:I CUBIC (Sender)I SACK (Receiver)I TCP IW5= 10 packetsI Size of the file sent:
I 9kBI 15MB
3Network Simulator4Bandwidth-Delay Product5Initial Window
Delay variation on Iridium 04/03/2020 14/40
NS-2 : one flow (2/12)
I Choice of the Iridium gateway:
Figure: Transfer time for random starting time and terminal positions
⇒ Hawaii’s gateway
Delay variation on Iridium 04/03/2020 15/40
NS-2 : one flow (3/12)
I Impact of the seam (n◦4)I Small file (9kB)
Figure: Transfer time simulations of a 9 kB file
⇒ New Delhi 124 ms outside the seam → 144 ms, an increase of 15.6%.⇒ LA 90 ms outside the seam → 179 ms, an increase of 98%.
Delay variation on Iridium 04/03/2020 16/40
NS-2 : one flow (4/12)
⇒ See what’s happening on the seam ?
0
2
4
6
8
10
41082.1 41082.2 41082.25 41082.3 41082.35 41082.40
2
4
6
8
10
Pack
et s
eque
nce
#
Time (s)
Dequeue timeEnqueue time
Reception time at 1st sat
Reception time at receiver node
Reception time of ACK
0
2
4
6
8
41082.1 41082.2 41082.2 41082.3 41082.3 41082.40
2
4
6
8
Pack
et s
eque
nce
#
Time (s)
Reception time at 1st satReception time at receiver node
Reception time of ACK
(a) 9kB file transfer throughthe seam
90
60
-30
-60
-900 30 60 90 120 150 180-30-60-90-120-150-180
unprojected longitude
S58
57
46
35
24
13
2
1
0
10
9
8
D
30
0
0 30 60 90 120 150 180-30-60-90-120-150-180
unpro
jecte
d latitu
de S
47
46
36
35
25
24
14
3
13
4
2
1
0
10
D
9
8
57
58 6
5
7
S
D
ACKs of pkts n°0,1ACKs of pkts n°2,3ACKs of pkts n°4..9
Terminal-Satellite link
Paths taken by:
-180 -150 -120 -90 -60 -30 30 60 90 120 150 1800
(b) Paths taken by ACK6packets duringtransmission
6ACKnowledgementDelay variation on Iridium 04/03/2020 17/40
NS-2 : one flow (5/12)
I Impact of the seam (n◦4)I Big file (15MB)
Figure: Transfer time simulations of a 15 MB file
⇒ No fatal impact on the duration of transfer. From which file size does it becomenoticeable ?
Delay variation on Iridium 04/03/2020 18/40
NS-2 : one flow (6/12)
Seam VS file sizes:
Figure: Impact of the seam on files’ transfers
⇒ Starting 500kB.
Delay variation on Iridium 04/03/2020 19/40
NS-2 : one flow (7/12)
I Impact of variation in delay due to sat’s elevation:
100
150
200210
1
0
0.2
0.4
0.6
0.8
1
1.2
1.41.5
12h21m 12h22m12s 12h23m24s 12h24m14s
CW
ND
(pac
kets
)
Rec
eivi
ng t
hro
ughput
(Mbps)
Time
Figure: Elevation Delay Variation
⇒ No impact (indeed the propagation delay varies smoothly within [2.6 ms ; 8.2 ms])
Delay variation on Iridium 04/03/2020 20/40
NS-2 : one flow (8/12)
I Impact of the delay variation due to intra-orbital handover:
100
150
200210
12h23m20s 12h24m18s 12h25m12s 12h25m50s
CW
ND
(packets
)
1 2
Time
0
0.2
0.4
0.6
0.8
1
1.2
1.41.5
Receiv
ing t
hroughput
(M
bps)
(a) Elevation tointra-orbital handover
100
150
200210
CW
ND
(packets
)
2 1
1
12h28m22s 12h29m24s 12h30m36s 12h31m48sTime
0
0.2
0.4
0.6
0.8
1
1.2
1.41.5
Receiv
ing t
hroughput
(M
bps)
(b) Intra-orbital toelevation handover
1→2 : delay increases, so packets and ACKs sent previously are receivedquickly, which induces the transmitter to increase the CWND7.2→1 : while for the reverse handover, it’s the opposite. Thus thethroughput goes from 1.5 Mbps to 1.36 Mbps (decrease of 7.48%).
7Congestion WiNDowDelay variation on Iridium 04/03/2020 21/40
NS-2 : one flow (9/12)
I Impact of the delay variation due to inter-orbital:
100
150
200210
12h56m12s 13h3m 13h9m 13h12m
CW
ND
(packets
)
1 3
Time
0
0.2
0.4
0.6
0.8
1
1.2
1.41.5
Receiv
ing t
hroughput
(M
bps)
(c) Elevation tointra-orbital handover
0
0.2
0.4
0.6
0.8
1
1.2
1.41.5
13h11m39s 13h12m18s 13h12m54s 13h13m19s
Receiv
ing t
hroughput
(M
bps)
3 1
Time
100
150
200210
CW
ND
(packets
)
100
150
200210
CW
ND
(packets
)
(d) Intra-orbital toelevation handover
1→3 : same explanation as before (1→2) the throughput decreases from1.47 Mbps to 1.35 Mbps (decrease of 8.16 %). Throughput and CWNDoscillation (the two peaks) because of intra-orbital handover. (2→1)3→1 : same explanation as before (2→1) the throughput decreases from1.47 Mbps to 1.35 Mbps (decrease of 8.16%).
Delay variation on Iridium 04/03/2020 22/40
NS-2 : one flow (10/12)
Settings:I Iridium:
I Plane : 6.I Sat. per plane : 11.I Up/Downlink bandwidth : 120 Mbps.I ISL bandwidth : 2 Gbps.I Buffer size : BDP
I TCP:I CUBIC (Sender)I SACK (Receiver)I TCP IW = 10 packetsI Size of the file sent:
I 9kBI 15MB
Delay variation on Iridium 04/03/2020 23/40
NS-2 : one flow (11/12)
Seam VS file sizes:
Figure: Impact of the seam on files’ transfers
⇒ No impact yet, need to push even further.
Delay variation on Iridium 04/03/2020 24/40
NS-2 : one flow (12/12)
Delay variation sources in higher rate:
Figure: Impact of the delay variation sources
Delay variation on Iridium 04/03/2020 25/40
Table of Contents
Context
Different intrinsic sources of delay variation
Simulations and ResultsNS-2 : one flowNS-2 : Multi-flow fairnessOpenSAND
Delay variation on Iridium 04/03/2020 26/40
NS-2 : Multi-flow : Test Architecture (1/7)
Figure: Example test Architecture for Low Throughput use case
Delay variation on Iridium 04/03/2020 27/40
NS-2 : Multi-flow : Settings (2/7)
I Start time of simulation : 15000.0s
I End time of the simulation : 15305.0s (outside the seam)
I One reference flow : at Perth (Australia)
Settings Low Throughput High Throughput
ISL Bandwidth 25Mbps 850Mbps
Up/Downlink Bandwidth 1.5Mbps 51Mbps
Full-duplex link Bandwidth 4Mbps 136Mbps
Full-duplex link Delay 10ms 10ms
Buffer size BDP BDP
Table: Simulation settings
Delay variation on Iridium 04/03/2020 28/40
NS-2 : Multi-flow : Low Throughput (3/7)
I Same start timeI Same RTT8
Figure: Same RTT multiflows
8Round Trip Time (roughly = 2 x One-Way Delay)Delay variation on Iridium 04/03/2020 29/40
NS-2 : Multi-flow : Low Throughput (4/7)
I Same start timeI Different RTT (Las Vegas < Ottawa < Perth)
(a) 2 flows share Perth VS LasVegas
(b) 2 flows share Perth VSOttawa
Delay variation on Iridium 04/03/2020 30/40
NS-2 : Multi-flow : Low Throughput (5/7)
I Same start timeI Different RTT : Jain’s Fairness index
(c) 2 flows jain’s index PerthVS Las Vegas handover
(d) 2 flows jain’s index PerthVS Ottawa handover
Delay variation on Iridium 04/03/2020 31/40
NS-2 : Multi-flow : Low Throughput (6/7)
I Different start time : Second flow is differed with 50s (afterthe non-permanent phase : Slow Start)I Same RTT
Figure: Same RTT multiflows different start time
Delay variation on Iridium 04/03/2020 32/40
NS-2 : Multi-flow : Low Throughput (7/7)
I Different start timeI Different RTT
(a) 2 flows share Perth VS LasVegas 50 s differed
(b) 2 flows share Perth VSOttawa 50 s differed
Delay variation on Iridium 04/03/2020 33/40
NS-2 : Multi-flow : High Throughput
I Same start timeI Same RTT : Same behaviorI Different RTT : Same behavior
I Different start timeI Same RTT : Same behaviorI Different RTT : Different behavior
(c) 2 flows share Perth VS LasVegas 50 s differed
(d) 2 flows share Perth VSOttawa 50 s differed
Delay variation on Iridium 04/03/2020 34/40
Table of Contents
Context
Different intrinsic sources of delay variation
Simulations and ResultsNS-2 : one flowNS-2 : Multi-flow fairnessOpenSAND
Delay variation on Iridium 04/03/2020 35/40
OpenSAND (1/4)
What is Amal up to now ?
Open Satellite Network Demonstrator is a satellite testbedemulating the main features of a SATCOM system maintained bythe collaboration work of Viveris Technologies, Thales AleniaSpace and CNES.
Delay variation on Iridium 04/03/2020 36/40
OpenSAND (3/4)
One Flow :I Low Throughput :
I Up/Downlink bandwidth : 1.5 MbpsI ISL bandwidth : 25 MbpsI Buffer size : BDP
Figure: Received throughput VS OWD
⇒ Currently under investigation ..Delay variation on Iridium 04/03/2020 38/40
OpenSAND (4/4)
One Flow :I High Throughput :
I Up/Downlink bandwidth : 49.5 MbpsI ISL bandwidth : 825 MbpsI Buffer size : BDP
Figure: Received throughput VS OWD
⇒ Currently under investigation ..Delay variation on Iridium 04/03/2020 39/40