Copyright by Burt H. Liebowitz 2004 IP NETWORKING OVER SATELLITE Burt H. Liebowitz May 2004 Global Internet Throughput vs. Round Trip Delay with Window Size as a Parameter 0 200 400 600 800 1000 1200 1400 1600 1800 2000 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Round Trip Delay in Seconds Effective Throughput in Kbps 4096 8192 16384 32768 65536 window size in bytes For Government, Military and Commercial Enterprises
This three-day course is designed for satellite engineers and managers in government and industry who need to increase their understanding of the Internet and how Internet Protocols (IP) can be used to transmit data and voice over satellites. IP has become the worldwide standard for data communications. Satellites extend the reach of the Internet and Intranets. Satellites deliver multicast content efficiently anywhere in the world. With these benefits come challenges. Satellite delay and bit errors can impact performance. Satellite links must be integrated with terrestrial networks. Space segment is expensive; there are routing and security issues. This course explains the techniques and architectures used to mitigate these challenges. Quantitative techniques for understanding throughput and response time are presented. System diagrams describe the satellite/terrestrial interface. The course notes provide an up-to-date reference. An extensive bibliography is supplied.
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Copyright by Burt H. Liebowitz 2004
IP NETWORKING OVERSATELLITE
Burt H. LiebowitzMay 2004
Global Internet
Throughput vs. Round Trip Delay with Window Size as a Parameter
0
200
400
600
800
1000
1200
1400
1600
1800
2000
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Round Trip Delay in Seconds
Eff
ect
ive
Th
rou
gh
pu
t in
Kb
ps
4096
8192
16384
32768
65536
window sizein bytes
For Government, Military and Commercial Enterprises
Copyright by Burt H. Liebowitz 2004
Seminar Outline1- Introduction and Purpose2- Fundamentals of Data Networking3- The Internet and Its Protocols4- Quality of Service Issues in IP Networks5- Satellite Data Networking Architecture6- System Design and Economic Issues for Satellite-
Based IP Networks7- TDMA/DAMA Design Example8- Predicting Performance in Mission-Critical Networks9- Conclusions and a View of the FutureBibliography and Table of Acronyms
Copyright by Burt H. Liebowitz 2004
Part 2 - Fundamentalsof Data Networking
Sender
Receiver
Network cloud
Application
Link
Application
Application
Host
Satellite node
HostNode
Node
Node
Node
Node
SourceHost
Host Host Destination
Host
Destination
Host
Satellite node
Application
Application
Host
• Overview
• Issues
• Protocol Layers
• Link Layer Protocols– Frame Relay
– ATM
– Aloha
– DVB
– Ethernet
• The Physical Layer
Copyright by Burt H. Liebowitz 2004
ATM Switching
NetworkHeader
2 bit Payload Typereserved
1 bit priority
8 4 4 8 4 2 1 1 8Header error
control - detects2 bit error; corrects 1
bit error
VPI VPI VCI VCI VCI
EndPoint
EndPoint
These VCs are switched under onevirtual path identifier
Copyright by Burt H. Liebowitz 2004
Digital Video Broadcast(DVB)
• A world-wide standardfor one-way transmissionof digital TV via satellite(S), cable (C) orterrestrial (T).
• Utilizes MPEG-2compression and packetstandard – fixed size188-byte cells
• Supports data as well asvideo transmissions.
• Supports multipleprogram streams, each ofwhich can be encrypted
• Will discuss in detaillater in the seminar
Satellite dish
Hub LAN
There is areturn linkstandardcalledDVB/RCS
Corporate and ContentProvider Hosts
DVB Hub
Virtual Circuit,Defined by aProgramIdentifier)
Copyright by Burt H. Liebowitz 2004
pop
Internet Structure
IP Network
IP Network
IP Network
Router
pop poppop
rasras
dasdas
NAP
pop
rasras
AS - Autonomous System
DNS - Domain Name Server
POP - Point of Presence
DAS - Direct Access System
RAS - Remote Access System
NAP - Network Access Point
HostY
(partof X)
Web Client
DNS
Client terminal contains browser toconnect to world-wide-web
Internet ServiceProvider (ISP)
AS 1
AS 3
AS 2
Network X
RouteAdvertisement:Network X ispart of AS1
Copyright by Burt H. Liebowitz 2004
Impact of TCP Window Size andRTT on Throughput
Throughput vs. Round Trip Delay with Window Size as a Parameter
0
200
400
600
800
1000
1200
1400
1600
1800
2000
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Round Trip Delay in Seconds
Effect
ive T
hro
ughput in
Kbps
4096
8192
16384
32768
65536
window sizein bytes
Effective Throughput in a Non-Congested, Error-Free Link with Channel Speed of2048 Kbps – assuming very long file transfer
Throughput vs. RTT With Window Size as a Parameter
RTT in seconds
Copyright by Burt H. Liebowitz 2004
Impact of Bit Errors on TCP/IPThroughput
File Size is I MB
RTT is 540 ms
Figure showsimpact of WindowSize (in Kbytes)
It should be notedthat impact of biterrors willdiminish as filesize decreases
• Unfortunately TCP cannot tell the difference between a packet loss due to bit errors or congestion -therefore TCP overreacts to bit errors by reducing throughput
• The moral of this story is that it is essential to have a link with a low bit error rate!
-8 -7 -6 -5
Bit Error Rate (10-n)
Effect of Bit Errors on Throughput
0
100
200
300
400
500
600
-8 -7 -6 -5
Bit Error Rate (10-n)
Thr
ough
put
in K
bps
8
32
64
Copyright by Burt H. Liebowitz 2004
VOIP Integration with IP Networks• Many of the quality issues can be resolved if
VOIP is confined to private networks– steps can be taken to insure that adequate
bandwidth is available to minimize delay, jitter andlost packets.
• This is accomplished by providing ample bandwidthand admissions control (H.323, SIP)
• Integration with PSTN and admissions control viaVOIP Gateways (for both public and private IPNetworks)– digitize, compress, admission control, signaling
There is an advantage to saturating a transponder - that is, to putup a single carrier and eliminate backoff. Can get more effectivepower and could reduce separation factor
Copyright by Burt H. Liebowitz 2004
A Shared Carrier Point to MultipointNetwork
Router
This is also a one to three network;however the carrier uplinked from the hubsite is seen by all the remote sites. Thecarrier is multiplexed. Each packet on thecarrier is identified with a destinationaddress. The appropriate destination (s)select the packet; the others discard thepacket
Shared Carrier Downlink
These are actuallythe same carrier,seen identically by allreceiving stations
Hub site
M
MM Router
MM Router
MM Router
DD
MM
DD
IPNetworks
Copyright by Burt H. Liebowitz 2004
TDMA/DAMA Return LinksEach site
transmits a burston the same
carrier at specifiedtimes - there canbe no collisions
Combineddownlink to hub
Time Division Multiple Access (TDMA)
Demand Assigned Multiple Access DAMA)
Copyright by Burt H. Liebowitz 2004
RequestingBandwidth - continued
• Timing Considerations
Tf Tf
Tsd Tsd
Tf
t1 - the time atwhich need for
more bandwidtharises
t2 - the time atwhich theterminal
acquires thenew bandwidth
Taq
Tf Time duration of frame
Tsd One way satellite delay
Taq Time to acquire newbandwidth
Tc - time tocollect requestsat master site
TfTaq ~ 2*Tsd + 3/2* +Tc
Assumption: requests are madeusing low utilization aloha signalingchannel - probability of collision isvery low.
ViaSat LinkStar DVB TDMA/DAMA Will soon have a DVB/RCS optionViaSat Surfbeam TDM –
DOCSISTDMA-DOCSIS Uses cable modem standard – not yet available:
designed for consumer market
Copyright by Burt H. Liebowitz 2004
TDMA DAMA Full MeshedNetworks
Similar to TDMA return link except that all remotesites look at the same down link; in that way eachremote site can transmit directly to each other remotesite without having to go thru a hub site
5
One site isdesignated as acontrol site; itprocessesbandwidthrequests andcreates theburst plan
12 3
4 6
Remote bursts are combined insatellite and transmitted on acommon downlink frequency
Copyright by Burt H. Liebowitz 2004
Satellite Adaptation for ReliableMulticast
• Use UDP for downstream
• Receiver sends periodic acks to indicate that some large number of blockshave been successfully received
• When bad block is received– Receiver picks random number to determine when it can send nack
• this is done to prevent all receivers from nacking a block sent in error from the hub
– While waiting to nack, receiver will process new blocks from the sender• during this time, if receiver sees retransmit of bad block, it updates its file and aborts the
nack
• if timer runs out before receiver sees a retransmit, the receiver transmits its nack
12345
Site 1 nacks firstSite 2 aborts nackSite 3 nacks Site 4 aborts nackSite 5 aborts nack
Second nack is ignored
4th block has error on uplink First nack causes sender to retransmit block 4
Hub
There are some scalabilityissues with this approach
Copyright by Burt H. Liebowitz 2004
A TDMA/DAMA Design Example
• We wish to design a 25-node, hub-based network
• We know the downstream and upstream data and voicerequirements
• We have three architectural alternatives– SCPC – MF/SCPC
– DVB – DVB/RCS
– TDM – MF/TDMA/DAMA (proprietary)
• Our task is to find the most cost-effective solution,considering– Capital costs
• amortized over five years
– Satellite bandwidth costs• based on a monthly lease over a five year period
Copyright by Burt H. Liebowitz 2004
NETWORK MODEL
Satellite dish
PRIVATENETWORK
R
HUB
INTERNET
25 Sites
• Major Applications– Access to Enterprise Data Base– Specialized Transactions– Email– Web Access– Voice
• 1 Theater
• 25 Sites per Theater ofOperation
• 1 hub per Theater
E
C: call managerE: encryptorF: firewallM: satellite TDM or TDMA modemR: routerV: voice over IP Gateway
V
E M
C
M E R
V C
F
EE
Copyright by Burt H. Liebowitz 2004
Impact of Number of Busy Sites
Satellite Bandwidth as a Function of Busy Sites
0
2000
4000
6000
8000
10000
12000
14000
16000
0 5 10 15 20 25
Number of Busy Sites
Sat
elli
te B
W i
n K
hz
SCPC-SCPC
DVB-RCS
Proprietary
Parameters• 25 sites total• Basic VOIP
– No compression, VAD orencryption
Total Cost Based on Number of Busy Sites
$-
$1,000
$2,000
$3,000
$4,000
$5,000
$6,000
$7,000
$8,000
0 5 10 15 20 25
Number of Busy Sites
Co
st i
n $
K SCPC-SCPC
DVB-RCS
Proprietary
Copyright by Burt H. Liebowitz 2004
Part 8 – Predicting Performance inMission-Critical Networks
• Overview and Definitions
• A Reference Problem
• Introduction to Queuing Theory– single server
– priority queues
• Application of Queuing Theory to Reference Problem
• Use of Simulation
Copyright by Burt H. Liebowitz 2004
A Closer Look at the System
• A call center with an operational outpost in an underservedregion of the world (“the remote site”)
• The nominal number of Operators is 192
• Operators answer calls– access a data base at the hub or at company locations (via an intranet)
• Average time between calls per operator is 120 seconds
• Response time is critical, as is voice quality
• We need to know if proposed satellite link can support operatorneeds– If so, can we support even more operators on the link?
Copyright by Burt H. Liebowitz 2004
R esp o n se an d D elay T im es
0
1
2
3
4
5
6
7
8
9
10
0 50 100 150 200 250 300
N um be r of C all Take rs
Res
po
nse
Tim
e in
sec
. fo
r P
rio
rity
2;
Per
use
r K
bp
s fo
r P
rio
r. 3
0.000
0.050
0.100
0.150
0.200
0.250
0.300
0.350
0.400
0.450
0.500
De
lay
Tim
e f
or
Vo
ice
Pa
ck
ets
(sec
on
ds)
average delay for c rit ic al trans ac tionbw for non-c rit ic al trans ac tionsaverage delay for voic eperc entile for priority voic eperc entile for non-priority voic e
RESPONSE TIME CURVES: We would like to know the average response time for eachof the transactions, based on the number of call takers serviced on the 1.8 Mbps line. In this
way we can see the impact of traffic on response time, and determine if we can indeedhandle the 192 call takers envisioned for this service.
Note: This graph is for traffic flowing from the hub -there will be less traffic in the other direction
192 CallTakers
Copyright by Burt H. Liebowitz 2004
Part 9 - A View of the Future
• Satellite Enhancements– More power and large antennas
– Spot beams and frequency reuse
– On board processing
– Inter-satellite links
– Ka and higher band satellites
• Advanced Satellites– Commercial
– Military
• The “Ideal” Earth Station
Copyright by Burt H. Liebowitz 2004
Frequency Reuse can ExpandCapacity of Satellite
• In this example we dividefrequency spectrum into 7segments– Frequency reuse factor is 7
• Assume– 100 beams
– 500 MHz spectrum
• Available frequency =
2*100*500/7
= 14.3 GHz
This factor of 2 comes from antenna polarization
31
2
45
6
73
1
2
45
6
7
31
2
45
6
73
1
2
45
6
7
Copyright by Burt H. Liebowitz 2004
Baseband Switching
Issues• Complexity
• Takes up weight and power
• Counter productive if too muchmulticast
• Layer 2 or Layer 3 switching?
Switch/Router
demod
demod
receiver
receiver
mod
mod
trans
trans
Benefits• Connectivity for spot beams
– Point to point
– Multicast
• 3 dB gain in link budget– Separation of up and down links
Optional PEP
Optional DAMA
Copyright by Burt H. Liebowitz 2004
MILITARY DATA NETWORKING SATELLITES- FUTURE
SYSTEM CAPACITY ON BOARD PROCESSING
INTER-SATELLITE LINKS
FREQUENCY BANDS
Wideband Gapfiller Satellite (WGS) First satellite to be launched in 2004
39, 125 Mhz channels preliminary estimate of 2.4 Gbps capacity per satellite
Yes – Channelizer divides capacity into 1872, 2.6 MHz subchannels. Channels can be switched based on network management commands
No X and Ka band (cross-banded)
References: www.losangeles.af.mil/smc/mc/wgs.htm; “Future US Military Satellite Communications Systems"” Glen Elfers and Stephen B. Miller, The Aerospace Corporation www.aero.org/publications/crosslink/winter2002/08.html Advanced EHF Under contract
Gigabits per second capacity Up to 8 Mbps per individual terminal. More than 12 times capacity of Milstar in some scenarios
RF switching Yes, will employ RF cross links
EHF bands because of limited availability of X and Ka for military use
References: “Future US Military Satellite Communications Systems"” Glen Elfers and Stephen B. Miller, The Aerospace Corporation www.aero.org/publications/crosslink/winter2002/08.html
Transformational Satellite (TSAT): EHF communications (44Ghz up; 20Ghz down), Space-based IProuter, .8 to 3.1 Gbps per satellite; Laser Cross Links, Ka and X-band payloads – “Implementing theGlobal Information Grid (GIG), Dr. Michael S. Frankel, DASD
Copyright by Burt H. Liebowitz 2004
Desirable Attributes for FutureMilitary Terminal
• MF/TDMA/DAMA– Bandwidth on demand with priorities– Can support either star or full mesh– Bandwidth efficient burst allocation
• 100s of terminals per network• Multi-beam support• Efficient bw utilization
– Turbocode or better– Dynamic coding and modulation
(DCM)
• Wide range of coding and modulation– BPSK, QPSK, 8PSK, 16QAM/PSK– Rate ½-8/9
• Multi frequency band– C,Ku,X,Ka
• Uplink power control
• MF/TDMA/DAMA– Bandwidth on demand with priorities– Can support either star or full mesh– Bandwidth efficient burst allocation
• 100s of terminals per network• Multi-beam support• Efficient bw utilization
– Turbocode or better– Dynamic coding and modulation
(DCM)
• Wide range of coding and modulation– BPSK, QPSK, 8PSK, 16QAM/PSK– Rate ½-8/9
• Multi frequency band– C,Ku,X,Ka
• Uplink power control
• L-band IF interface• Remote site can support up to 2 Mbps• Wide range of terrestrial interfaces• Network Management channel• Router functionality
– Basic routing functions– High degree of QoS including
• MLPP (Multilevel Precedence andPreemption)
– Header Compression
• Built in TCP performanceenhancement
• High level of Security– Large keys, AES and 3DES– Authentication
• L-band IF interface• Remote site can support up to 2 Mbps• Wide range of terrestrial interfaces• Network Management channel• Router functionality
– Basic routing functions– High degree of QoS including
• MLPP (Multilevel Precedence andPreemption)
– Header Compression
• Built in TCP performanceenhancement
• High level of Security– Large keys, AES and 3DES– Authentication
REFERENCE: “IP Modem Functional Capabilities Description” (FCD) Version 1.5April 15, 2004, Defense Information Systems Agency (DISA)