Asif Iqbal

1
Optical Burst Switching and Hierarchical Optical
Burst Switching

• Asif Iqbal
• Interconnection Networks (Fall 2004)
• Kent State University

2
Importance

• The importance of Optical Burst switching can be
  noticed by evaluating the drawbacks of Optical
  packet switching
• In general, OPS requires optical buffering and
  line-rate header parsing (what and why?)
• There are two types of Optical Packet Switching

3
Importance (cont)

• In synchronous optical packet switching (slotted
  network), packets have the same size.
• The arriving packets must be aligned with the
  local clock (not very simple to do for high speed
  traffic)

4
Importance (cont)

• In asynchronous optical packet switching
  (un-slotted network), packets have different
  sizes
• Since packets have unpredictable arrival time on
  the switch, there is increased probability for
  packet contention
• Increased packet loss ratio
• Literature references the use of TOWC in FDL for
  contention resolution

5
Definitions

• Optical Burst Switching a transport technology
  of transmitting data in the form of bursts in an
  all optical buffer-less network
• The idea is to separate control plane
  (electrical) from data plane (optical)
• Makes the network transparent to the content of
  the burst

6
Definitions (cont)

• Burst continuous set of data bytes or packets
• Variable length

7
Design Requirements

• Fast optical switches
• Low signal degradation
• Large switching matrices
• Non-blocking architecture
• Use of wavelength switching WDM and wavelength
  converters
• Edge-core architecture for burstification

8
Design fundamentals

• Bandwidth is reserved before hand, consumed
  released
• Each data burst travels through an all optical
  path between source and destination
• No Opto-electric or Electro-Optical conversion is
  needed
• A burst control packet (BCP) precedes the data
  burst to reserve resources in the switches and
  channel along the path
• Each control packet is processed electronically
  to setup optical switch/cross-connect
  electronically
• The delay between sending BCP and data is called
  offset time

9
Design fundamentals (cont)
10
Variations

• There are 4 main design variations
• Tell-and-Wait (TAW)
• Source sends a control packet along the path to
  inform that it wants to transmit a burst
• If all intermediate switches accommodate request
  accepted, source transmits
• Otherwise request is refused, and source tries
  again

11
Variations (cont)

• Tell and Go (TAG)
• Source first sends a control packet on a separate
  control channel to reserve bandwidth
• After the burst, source sends a control packet to
  release the bandwidth
• A loss of the tear down signal can result in
  bandwidth wastage

12
Variations (cont)

• Use of an in band terminator (IBT)
• Each burst has a header and a delimiter to
  indicate beginning and end of burst
• Uses virtual cut through (what and why?) instead
  of store and forward

13
Variations (cont)

• Reserve a fixed duration (RFD)
• Like Tell and Go, a control packet is sent before
  the burst
• However in this case, the control packet contains
  the length of the burst

14
Variations (cont)

• RFD based designs
• Just enough Time (JET)
• A source sends out a control packet (similar to a
  set-up request), which is followed by a burst
  after some offset time
• A variation of Jet is pJet where bursts are
  classified in to multiple classes to provide
  differentiated service

15
Variations (cont)

• In pJet
• More priority classes get an extra offset time
  which allows a control packet to make bandwidth
  reservation in much more advance thus giving a
  greater chance of success
• In both RFD based Jet and pJet, control packet is
  required to specify the duration of the following
  burst

16
Variations (cont)

• RFD based designs are more attractive
• Uses an offset time approach
• This time offset must be large enough to process
  the header and setup the switching matrix for the
  data bursts
• The number of the traversed nodes must be known
  at the edge node and considered to determine the
  burst delay
• No buffer is needed on intermediate nodes because
  of the offset time approach (2 x one way delay
  plus processing delay)

17
Variations (Cont)

• Will it not be a better solution to send header
  and data bursts together and only delay the data
  burst before it enters the switching matrix?

18
Hierarchical OBS

• A situation may arise when a control packet
  reaches a node and fails to reserve the necessary
  bandwidth because another burst is scheduled for
  the same time.
• Normally in OBS, the second burst is simply
  dropped
• In HBOS, the second burst is rescheduled after
  the first burst has completed transmission
• The second control packet is reset with a new
  scheduled time of the corresponding burst before
  being forwarded to the next hop
• When the second burst arrives, it is delayed by
  the amount of time indicated by its modified
  control packet

19
OBS versus HOBS

• Considering bandwidth utilization and packet drop
  probability, Patankar, N. Arora and S. Asthana
  showed that
• As offset time increases, burst loss ratio
  decreases in HOBS compared to OBS (Why?)
• Bandwidth utilization increases in both OBS and
  HOBS in similar way with time

20
OBS and QoS

• Application layer QoS done on the edge during
  burstification and scheduling
• Physical layer QoS function of optical
  components, fiber etc.
• QoS and priority pre-emption during congestion
  (why not an issue in HOBS?)

21
References (Literature Review)

• David Q. Liu and Ming T. Liu, Differentiated
  Services and scheduling scheme in Optical Burst
  Switched Networks, June 2002.
• D. Q. Liu and M. Liu, Priority-based burst
  scheduling scheme and modeling in optical
  burst-switched WDM networks, In proceedings of
  International Conference on Telecommunications
  ICT 2002, June 2002.
• Jonathan Turner, Terabit burst switching,
  Journal of High Speed Networks, 83-16, 1999.
• C. Qiao and M. Yoo, Choices, Features and Issues
  in Optical Burst Switching, June 1999.
• A. Patankar, N. Arora and S. Asthana, Optical
  Burst Switching in Hierarchical Networks,
  November 1999.
• Hao Buchta, Erwin Patzak, jurgen Saniter,
  Christoph Gauger, Maximal and Effective
  throughput of Optical Switching Nodes for Optical
  Burst Switching, July 1998.
• Jeyashankher Ramamirtham, Jonathan Turner, Time
  Sliced Optical Burst Switching, IEEE Infocom,
  0-7803-7753-2/03, 2003.
• Hiroaki Harai, Masayuki Murata and Hideo
  Miyahara, Performance of Alternate Routing
  Methods in All-Optical Switching Networks, IEEE
  Infocom, 0-8186-7780-5/97, 1997.
• Rajagopal Kannan, Sibabrata Ray, Radim Bartos,
  An Optical Switching Architecture for
  Hiearchical Group Communication, DARPA grant no.
  F30602-02-1-0198, LEQSF grant no.
  (01-03)-RD-A-06, January 2003.
• Chiaro Networks, Discussing Optical Phased Array
  Technology for High-Speed Switching, A Technical
  Paper, 111802.01.2, 2003.