TCP/IP Over Satellites

1
TCP/IP Over Satellites
Kofi Weusi-Puryear Sridhar Dhulipala Sanjay
Jain Tejaswi Redkar
SJSU, CMPE 206, Spring 1998
2
Why TCP/IP over satellite?

• Ease of installation, remote accessibility, and
  low cost of running/maintenance.
• Useful for areas or countries that have little
  terrestrial infrastructure.
• Mobility of the devices ease of adding/removing
  subsystems
• Applications
• Internet connectivity.
• Telecom industries (pagers, cell phone, wireless
  modem..)
• Defense

3
Disadvantages

• Long delay, especially in geosynchronous systems
• High probability of bit errors.

Objectives

• Highlight TCP/IP performance issues with
  improvements/suggestions
• The following mechanisms(current and proposed are
  discussed)
• Path-Maximum Transmission Unit (MTU) discovery
  and Forward Error Correction (FEC)
• Selective Acknowledgements
• Congestion Control using slow start, congestion
  avoidance, fast transmit and fast recovery
  algorithms.
• TCP window size.

4
Path-MTU discovery

• A non-TCP mechanism, uses Internet Control
  Message Protocol (ICMP)
• MTU limit on the maximum datagram size on a
  network media.
• Path MTU discovery Maximum packet size on a
  network path without IP packet fragmentation.
• Allows TCP to use the largest possible packet
  size, without incurring the cost of fragmentation
  and reassembly.
• Disadvantage may cause a long pause before TCP
  is able to start sending data.
• In practice, not very time consuming due to wide
  support of common MTU values.

5
Forward Error Correction(FEC)

• Sender adds repair data to a media stream so
  packet loss can be repaired by the receiver.
• Repair data can be media specific or media
  independent.
• Media specific Employs knowledge of a media
  compression scheme.
• Low latency. Only a single-packet delay being
  added. Suitable for interactive applications
  which require low end to end delays.
• - Complex encoding/decoding slows sender
  receiver
• Media independent
• Uses parity based FEC, simplified
  encoding/decoding
• - High latency (due to more lost packets).

6
Selective Acknowledgements (SACKs)

• Recently approved extension to TCP.
• Makes it possible for TCP to acknowledge data
  received out of order.
• Previously TCP had only been able to acknowledge
  data received in order. This could lead to
  needless retransmissions, in case the sender
  transmits out of order.
• Improves the efficiency of TCP retransmissions by
  reducing the retransmission period.
• Helps TCP better evaluate the available path
  bandwidth in a period of successive losses and
  avoids doing a slow start.

7
TCP Congestion Control Mechanisms

• TCPs start a connection with the sender injecting
  multiple segments into the network, up to the
  window size advertised by the receiver.
• Problem with this Approach ?
• Congestion at an intermediate router (which
  causes packets to be discarded) can reduce the
  throughput of a TCP connection drastically.
• To avoid generating an inappropriate amount of
  network traffic TCP employs four congestion
  control mechanisms. These algorithms are slow
  start, congestion avoidance, fast retransmit and
  fast recovery
• TCP uses two variables to accomplish congestion
  control.
• The first variable is the congestion window
  (cwnd) The amount of data the sender can inject
  into the network.
• The second variable is the slow start threshold
  (ssthresh) which is used to decide which
  algorithm is to be used.

8
Slow Start Algorithm

• When a host begins sending data on the network it
  is required to use the slow start algorithm at
  the beginning of the transfer.
• Slow start begins by initializing cwnd to 1
  segment (i.e., the segment size announced by the
  other end, or the default, typically 536 or 512
  bytes).
• This forces TCP to transmit one segment and wait
  for the corresponding ACK.
• For each ACK that is received, the value of cwnd
  is increased by 1 segment.
• For Example after the first ACK is received the
  cwnd will be 2 segments and the sender will be
  allowed to transmit 2 data packets
• This continues until either the cwnd reaches the
  maximum window size or packet loss is detected.

9
Congestion Avoidance

• Congestion avoidance algorithm deals with packet
  loss which occurs due to congestion of the
  network.
• In the congestion avoidance mode the cwnd is
  increased only one segment at a time until the
  max window size is reached.
• When congestion occurs (indicated by a timeout or
  the reception of duplicate ACKs), one-half of the
  current window size is saved in into the ssthresh
  .
• If cwnd lt ssthresh, TCP is in slow start
  otherwise TCP is performing congestion avoidance.
• Congestion avoidance and slow start are
  independent algorithms with different objectives,
  but in practice are always implemented together.
  The following diagram explains how the algorithms
  work together.

10
Slow Start and Congestion Avoidance combined
implementation.
11
Slow Start and Congestion Avoidance (Summary)

• The TCP protocol reacts to packet loss by using
  Slow Start and Congestion Avoidance because it
  assumes there is network congestion.
• The above method works well for shared networks,
  but not for wireless networks (for example
  Satellite links) where the cause of the problem
  could be due to bit errors. The solution for bit
  errors is retransmission.
• Therefore the Slow Start and Congestion Avoidance
  Algorithms do not enhance the TCP performance
  over Satellite links, but are required in case of
  a network collapse.

12
Fast Retransmit and Fast Recovery

• Found in many Unix variants of TCP for several
  years.
• Not documented as a Standards Track TCP mechanism
  until Stevens in 1997 (RFC 2001).
• TCP Reno does use Fast Retransmit and Fast
  Recovery.

13
Fast Retransmit

• TCP's default mechanism to detect dropped
  segments is the retransmission timeout (RTO)
  based on observations of the Round-Trip Timeout.
• TCP ACKs always acknowledge the highest in-order
  segment that has arrived. If a segment arrives
  out-of-order the ACK triggered will be for the
  highest in-order segment, rather than the segment
  that just arrived. Thus the receiver is going to
  send a duplicate ACK if a segment is lost.

14
Fast Retransmit (continued)

• The fast retransmit algorithm uses these
  duplicate ACKs to detect lost segments.
• If 3 duplicate ACKs arrive at the data
  originator, TCP assumes that a segment has been
  lost and retransmits the missing segment without
  waiting for the RTO to expire.
• Thus Fast Retransmit reduces the time it takes a
  TCP sender to detect (and react to) a single
  dropped segment.

15
Fast Recovery

• After a segment is resent using fast retransmit,
  the fast recovery algorithm is used to adjust the
  congestion window.
• Fast recovery halves the segment sending rate and
  begins congestion avoidance phase immediately,
  skipping slow start.
• By skipping slow start Fast Recovery reduces
  wasting satellite pipe bandwidth.

16
Fast Retransmit and Recovery (continued)

• When multiple segments are lost in a given window
  of data, only one of the segments will be resent
  using Fast Retransmit. The rest of the dropped
  segments must wait for the RTO to expire, which
  causes TCP to revert to slow start.
• TCP knows that packets are still flowing through
  the network when retransmitting due to duplicate
  ACKs. However, when resending a packet due to a
  the expiration of the RTO, TCP cannot infer
  anything about the state of the network and
  therefore must proceed conservatively by using
  the slow start algorithm.

17
Delay and Window Size

• One of the limiting factors for throughput is the
  delay of a connection.
• TCP is designed as a sliding window protocol,
  hence, a fixed maximum amount of data (receive
  window size) is on the link between sender and
  receiver as shown in the figure below.

• As a consequence the maximum throughput is
  limited by the Round Trip Time (RTT) of the link
  
• throughput(max) receive buffer size/round trip
  time
• If this equation is rearranged to
• round trip time x throughput lt receive buffer
  size
• one will get the condition for the delay
  bandwidth product (DBP) that is often used to
  characterize connections.

18
Delay and Window Size (examples)

• For a terrestrial connection with 8 KBytes window
  size and 10ms RTT this leads to
• throughput(max) 8KBytes/10ms 800,000
  Bytes/s.
• For a satellite connection with 64 KBytes window
  size maximum in standard TCP) and 560 ms RTT
  follows
• throughput(max) 64KBytes/560ms 115,00
  Bytes/s
• As a result, at least a 100 KByte window is
  needed to fully utilize a T1-link (1.544 Mbps)
  with 560 ms RTT.

19
Solutions to Round-Trip Time Measurement (RTTM)
ProblemWindow Scale Option

• The window scale extension expands the definition
  of the TCP window to 32 bits and then uses a
  scale factor to carry this 32-bit value in the
  16-bit Window field of the TCP header (SEG.WND in
  RFC-793).
• The scale factor is carried in a new TCP option,
  Window Scale. This option is sent only in a SYN
  segment (a segment with the SYN bit on), hence
  the window scale is fixed in each direction when
  a connection is opened.
• The maximum receive window, and therefore the
  scale factor, is determined by the maximum
  receive buffer space.
• In a typical modern implementation, this maximum
  buffer space is set by default but can be
  overridden by a user program before a TCP
  connection is opened.
• This determines the scale factor, and therefore
  no new user interface is needed for window
  scaling.

20
Round-Trip Time Measurement (RTTM)

• Accurate and current RTT estimates are necessary
  to adapt to changing traffic conditions and to
  avoid an instability known as "congestion
  collapse" in a busy network.
• When the window size is tens or hundreds of
  packets, the RTT estimator may be seriously in
  error, resulting in spurious retransmissions.
• It is vitally important to use the RTTM mechanism
  with big windows otherwise, the door is opened
  to some dangerous instabilities due to aliasing.
  Furthermore, the option is probably useful for
  all TCP's, since it simplifies the sender.

21
PAWSProtect against Wrapped Sequence

• The PAWS Mechanism a simple mechanism to reject
  old duplicate segments that might corrupt an open
  TCP connection.
• PAWS operates within a single TCP connection,
  using state that is saved in the connection
  control block.
• PAWS uses the same TCP Timestamps option as the
  RTTM mechanism and assumes that every received
  TCP segment (including data and ACK segments)
  contains a timestamp SEG.TSval whose values are
  monotone non-decreasing in time.
• The basic idea is that a segment can be discarded
  as an old duplicate if it is received with a
  timestamp SEG.TSval less than some timestamp
  recently received on this connection.

22
Conclusions
Problems Long delay, especially in
geosynchronous systems Inefficient use of
satellite channel High probability of bit errors