on the impact of intrinsic delay variation sources on ...€¦ · i leo orbit, circular and nearly...

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


Table of Contents

Context


Simulations and Results


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.


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


Simulations and Results


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



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



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.



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



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.



90

60

30

0

-30

-60

-900 30 60 90 120 150 180-30-60-90-120-150-180


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




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



I Choice of the Iridium gateway:

Figure: Transfer time for random starting time and terminal positions

⇒ Hawaii’s gateway



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%.



⇒ 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


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


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 ?



Seam VS file sizes:

Figure: Impact of the seam on files’ transfers

⇒ Starting 500kB.



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])



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


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%).


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


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.


NS-2 : one flow (12/12)

Delay variation sources in higher rate:

Figure: Impact of the delay variation sources


Table of Contents

Context




NS-2 : Multi-flow : Test Architecture (1/7)

Figure: Example test Architecture for Low Throughput use case


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


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


I Same start timeI Different RTT (Las Vegas < Ottawa < Perth)

(a) 2 flows share Perth VS LasVegas

(b) 2 flows share Perth VSOttawa



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



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



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


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


Table of Contents

Context




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.


OpenSAND (2/4)

Figure: One-Way Delay


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

on the impact of intrinsic delay variation sources on ...€¦ · i leo orbit, circular and nearly...

Documents