Cross layer TCP-IP presentation on behalf of my cooperation with the University of Siena. Madrid 2006.

1. SatNEx II, 2nd ja2230 integration meeting, September 15, 2006, Madrid,
Spain
SatNEx II
Nico-Liberato Candio, Giovanni Giambene
CNIT - Research Unit of the University of Siena,
Via Roma, 56, 53100 Siena, Italy
"CROSS-layer optimization of transport
layer in SATellite networks" (CROSS-
SAT)
(ISTI, CNIT, DLR, ISTI, TUG, UoB)

2. SatNEx II, 2nd
ja2230 integration meeting, September 15, 2006, Madrid,
Spain
CROSS-SAT: objectives
● This activity focuses on TCP performance as a result of interactions
with lower layer protocols, and, in particular:
○ Channel conditions,
○ ACM (modulation & coding levels),
○ Radio resource management (e.g., DAMA schemes).
● We have interested to employ/compare different TCP versions as
well as TCP-friendly protocols.
● Our aim is to prove that the interaction/joint optimization of different
protocol layers can permit to improve the higher-layer goodput as
well as user satisfaction (QoS requirements).
● Current partners: ISTI, CNIT, DLR, TUG, UoB.

3. SatNEx II, 2nd
ja2230 integration meeting, September 15, 2006, Madrid,
Spain
CROSS-SAT: protocol stack
● The cross-layer approach is the study
of interactions of lower layers and
Internet levels in order to optimize
system performance.
● The ETSI TC-SES/BCM working
group has defined a layered
architecture where lower layers
depend on satellite implementation
and higher layers are those of the
Internet TCP/IP reference model
● These two independent stack
(satellite-dependent part and satellite-
independent part) are interconnected
via SI-SAP (Satellite Independent-
Satellite Access Point).

4. SatNEx II, 2nd
ja2230 integration meeting, September 15, 2006, Madrid,
Spain
Cross-layer method taxonomy
●Implicit cross-layer design:
○There is no exchange of information among different
layers during operation, but in the design phase cross-
layer interactions are taken into consideration for a joint
optimization.
●Explicit cross-layer design:
○Signaling interactions among (non-)adjacent protocol
levels are employed so that dynamic adaptation on the
basis of different protocol layer behaviors is possible.

5. SatNEx II, 2nd
ja2230 integration meeting, September 15, 2006, Madrid,
Spain
Explicit cross-layer
● Cross-layer information can be
exchanged from higher to lower
layers (top-down approach) or from
lower to higher layers (bottom-up
approach).
● The cross-layer approach allows
the direct exchange of control
information (signaling) even
between non-adjacent layers
● For instance, a 'get function' can be
used by higher layer protocols to
acquire the internal state of lower layer
protocols; moreover, a 'set function' can
be adopted by higher layer protocols to
change the state of lower layer
protocols.

6. SatNEx II, 2nd
ja2230 integration meeting, September 15, 2006, Madrid,
Spain
Scenario: main characteristics
●GEO satellite network
●Transparent satellite system
●Fixed users
●Use of the DVB-S2/RCS standard with ACM
●End-to-end system performance analysis
●TCP-based and UDP-based flows (TCP-
friendly protocols)

7. SatNEx II, 2nd
ja2230 integration meeting, September 15, 2006, Madrid,
Spain
Scenario: details
● We have implemented our
scenario under the ns–2
environment
● A GEO satellite bent pipe (using
SatNode_ class of simulator)
● Earth stations in LOS (Line of
Sight ) with the GEO satellite
● A memoryless channel with
uncorrelated losses
● FTP traffic during for all the length
of the simulation (1000 s)
● Different TCP techniques are
used: NewReno, SACK, Hybla
and Westwood+.
Network architetture

8. SatNEx II, 2nd
ja2230 integration meeting, September 15, 2006, Madrid,
Spain
Preliminary results:
● We are interested in defining an
optimal criterion for the mode
selection between mode on the
basis of the TCP goodput
performance
○ We have adopted two distinct
transmission modes that are operated at
parity of transmission bandwidth (2 MHz)
and power. In particular, we have:
○ Mode #1: BPSK with rate 1/2 convolutional
encoder, and resulting information bit-rate of 1
Mbit/s;
○ Mode #2: QPSK with rate 3/4 convolutional
encoder, and resulting information bit-rate of
3Mbit/s.
Mode #1
0 2 4 6 8 10 1
2Eb
/N0
[dB]
0
0.5
1
1.5
2
2.5
3
TCP goodput
[Mbit/s]
Opt. mode
transition
threshold for
Westwood+
Opt. mode
transition
threshold for
NewReno and
SACK
Opt. mode
transition
threshold for
Hybla
Mode #2
Hybl
a
Westwood+
SACK
NewRen
o

9. SatNEx II, 2nd
ja2230 integration meeting, September 15, 2006, Madrid,
Spain
Future plans
● Use of real channel traces and PHY performance evaluation or use
of BER traces (DLR) on the basis of radio channel and MODCOD
switching conditions. The reference is here on the proposed DVB-
S2 return channel
○ We need to cooperate with partners for having channel traces and/or
BER traces to be included in the ns-2 simulator
● Possible introduction of a DAMA scheme in the simulator to study
the impact of the transport level of PHY adaptivity and dynamic
bandwidth allocation.
● Possibility of real experimentation of the mode switching criterion on
a real satellite system (TUG).
● Extension of the study to the case of mobile users (need of
appropriate channel modeling).
● Focus on cross-layer issues for the elements interconnecting
satellite and terrestrial networks (PEP), evaluating differences

10. SatNEx II, 2nd
ja2230 integration meeting, September 15, 2006, Madrid,
Spain
Time plan
●Provision of BER traces on the basis of radio
channel and MODCOD switching conditions
(DLR, by the end of September).
●Definition of possible cross-layer experiments
(TUG, by mid November) to test MODCOM
switching criteria on the basis of transport layer
performance.
●Possible inclusion of a DAMA scheme (2007).

11. SatNEx II, 2nd ja2230 integration meeting, September 15, 2006, Madrid,
Spain
Any comments, proposal,
suggestions, interest ?
Thanks!