Research Methodology

1
Research Methodology

2
Doctor of Technology Thesis

• A Multi-Channel Network Architecture Based on
  Fast Circuit Switching
• by Per Lindgren

3
At the end you should have heard

• Optical Multi-Channel Networks
• Packet switched or circuit switched networks
• Understand about DTM
• New Interface design to cater for High bandwidth
  networks

4
About Per Lindgren
5
Rationale for this study

• Evolution of Fiber and Optical networks
• Increase of link Capacities
• Broadband networks vs Integration of Video, VoIP,
  and other distributed multimedia applications
• Integration of communication, computing and media

6
Issues addressed by this thesis

• Bottlenecks in processing and buffering in switch
  points
• Bottlenecks in network links (fiber), protocols
  and host interfaces
• Consequences of a communication bottleneck change
  ( I.e., circuit-switched versus packet-switched
  networks)

7
New Proposal

• New network architecture based on DTM
• New Host Interface

8
The thesis examines

• Optical multi-channel networks, eg. Based on WDM
• Believe that parallel channel structure in such
  networks is more cost effective solution than
  increasing the bit rate of a single bit stream.
• The interface between the network and the host
  interface and the host interface design

9
What technology at the time this thesis was
written

• ATM
• FDDI
• Fast Ethernet was springing

10
End to end communication
B
A
Network B
Network A
Switching Device
11
End to End Communication
Communication media
Interface design
Switch point
Routing protocol
Protocol
Protocol

• Communication media
• Interface design
• Protocol Design

12
Domain Scope of the Thesis

• Two different aspects of end-to-end
  communication
• Service requirements and Performance aspects,
  within two different areas
• Network Architecture
• Host Interface Design

13
Contents of the thesis
(179 pages)

• Introduction 8 pages
• Architecture for high Capacity Communications
• Background ..3
• Existing Network Architectures . 10
• Multichannel Network Systems .. 6
• Host Interface Design . 10
• DTM Circuit Switching Enhanced for Future
  Integrated Services
• Basic Principles of DTM 7
• DTM Protocol Suite . 30
• DTM Enhancements .. 6
• DTM Prototype Network .. 8

14
Contents of the thesis

• DTM Using Parallel Bitstreams
  35 pages
• Parallel DTM Protocol Aspects.
• Parallel DTM Hardware aspects .
• Parallel DTM Synchronization Aspects..
• Prototype Implementation
• Distributed Switching.
• Simulation of Parallel DTM.
• Simulation Results..
• DTM Host Interface.. 20
  pages
• The DTM Shared Memory Host Interface..
• A Low-Cost Interface to the DTM Network.
• Hardware Aspects of the Prototype Interface
• Performance Measurements ..
• Conclusion and Future Work ..33
  Pages
• References .. 16 pages
  with 205 ref

15
Packet Switching

• Internet Network

16
Circuit Switching

• Telephone Network

17
Competing Technologies

• ATM Asynchronous Transfer Mode
• Gigabit Ethernet
• FDDI Fiber Distributed Data Interface
• DPT Dynamic Packet Transport
• DTM Dynamic Synchronous Transfer Mode

18
Competing Technologies
19
ATM What is it? Pros? And Cons?

• Benefits
• Supports different types of data
• High transmission rates
• Good support for real-time traffic
• Scales easily
• Well suitable for long distances
• Drawbacks
• Complex technology
• Expensive
• Does not interoperate well with Ethernet
• No support for broadcast or multicast
• Only QoS for some ATM traffic classes
• Large amount of overhead

• Asynchronous Transfer Mode (ATM) is a packet
  switching and multiplexing technology designed as
  a general-purpose connection oriented (virtually
  circuit switched) transport mechanism for a wide
  range of services.

20
FDDI What is it? Pros? and Cons?

• FDDI uses a dual ring topology.
• During normal operation, the primary ring is used
  for data transmission, and the secondary ring
  remains idle.
• The primary purpose of the dual rings is to
  provide superior reliability and robustness.

• Benefits
• Reliable
• Built in redundancy
• Large MTU size
• Well tested
• Drawbacks
• Limited traffic capacity (100 Mbps)
• More expensive
• Lacks multicast support
• No future potential

21
Gigabit Ethernet- What is? Pros and Cons

• Ethernet standard 802.3 started in the 1973
• Improved to fast ethernet 100Mbps
• 1995 Began the Gigabit Ethernet standard 802.3z
• 10Gbps Standard 802.3ae on the way

• Benefits
• Cheap
• Well known
• Commonly used
• Simple to manage
• Support for multicast and broadcast
• Low overhead
• Drawbacks
• No Bandwidth guarantees
• Best effort Problems for applications such as
  VoIP, VC or any Real time application
• Small MTU size
• Limited management capability

22
DTM What is it? Pros? And Cons?

• Developed at KTH
• Built in circuit switching technology
• But utilizes the best features of Circuit and
  packet switching technologies
• It enables the user to set up and tear down the
  circuits instantly (like in a packet switched
  system)
• while getting reservation of bandwidth (which all
  circuit switched systems offer)

• Benefits
• Offers true QoS
• Dynamic increase/decrease of bandwidth (DTM
  channels)
• Fixed overhead size (depending on number of
  nodes), i.e. not proportional to link data speed.
  Small total overhead
• Fast set-up and tear-down of channels compared to
  other circuit switched technologies
• Simple and good support for multicast
• Compatible with other technologies, support for
  IP-routing, Ethernet bridging, VLAN and E1/T1
  tunneling. Can be run on dark fiber, DWDM and
  SDH/SONET (not all DTM versions)
• Deterministic delay
• High capacity

23
DTM What is it? Pros? And Cons?

• Drawbacks
• Circuit switched. Setting up a connection takes
  time (even though it is faster than other circuit
  switched alternatives), which is negative for
  short transfers
• New technology with an uncertain future
• Dynarc and Net Insight develop different
  protocols that are incompatible
• Minimum channel speed is 512 kbps, which might be
  unnecessary high for small transfers
• Complex, with unnecessarily many parameters
• Expensive,

24
DPT What is it? Pros? and Cons?

• Developed by CISCO
• Official name is Resilient Packet Ring - IEEE
  802.17
• DPT is designed to be an IP packet optimized
  transport solution.
• Combine the bandwidth efficiency and
  services-rich capabilities of IP routing with the
  self-healing capabilities of fiber rings.
• The key applications for DPT include cable data
  backbone access rings and regional transport
  rings.

• Benefits
• Fair allocation of bandwidth
• Low overhead
• Effective use of bandwidth (Spatial Reuse)
• Very Reliable and Robust
• Redundancy and large MTU size
• native multicast support and can distinguish
  between high priority and low priority packets
• Drawbacks
• Double ring needed
• Delays caused by buffering in nodes

25
Host Interface Design
CPU
RAM
System Bus
Bottlenecks
Host Interface
Transmission Medium
26
DTM Circuit Switching Enhanced for Future
Integrated Services

• DTM Prototype Network

• DTM Enhancements
• Fast Channel Creation
• Fast Channel Establishment over several hops
• Dynamic Signaling
• Base Frames consisting of multiple cycles
• Virtual Networks
• Slot Reuse in DTM
• Parallel DTM
• Switch Replication

                                                
    Fig 1. DTM Prototype Node
27
DTM Using Parallel Bit streams

• Two ways to increase transmission capacity
• Increase bit rate on the fiber
• Use parallel bit streams
• In both we can use
• Wavelength division multiplexing (WDM)
• Space Division Multiplexing (SDM)
• The proposed solution uses parallel bit streams
  with WDM

28
DTM Using Parallel Bitstreams

• Prototype Implementation
• The prototype uses replicated hardware to receive
  data on both wavelengths simultaneously.
• That is possible due to use of two demux, two
  slot counters and two channel tables on the
  board.
• Distributed Switching
• Distributed Switching scheme between two fibers
  in parallel DTM is done electrically.
• Is based on the shared medium to avoid switch
  bottleneck

• Simulation of Parallel
• Analyses were performed with regard to Slot
  Utilization, blocking and access delay for a dual
  bus with a varying number of parallel bit streams
• The effects of different user requirements,
  resource management schemes and traffic scenarios
  were also analysed.

29
DTM Using Parallel Bitstreams

• DTM Simulation Results
• It has been shown that Centralized Slot
  management schemes perform very well specially
  with partially-equipped nodes
• Fragmentation of the pool of free slots is
  normally very modest
• This scheme simplifies network management
• Disadvantages setup delay increases nodes are
  dependent on the slot server, which makes the
  system sensitive to failures

30
DTM Using Parallel Bitstreams

• DTM Simulation Results
• Distributed scheme has several benefits compared
  to the centralized.
• No risk of bottlenecks caused by a a single node
• Access delay is shorter, redundancy is higher
• Scales well to longer buses
• A Hybrid scheme is however likely to give the
  best overall performance. For instances
  distribute slots along nodes. Keep part in
  several slot servers that share the global state
  information.

31
DTM Host Interface

• DTM shared memory host interface
• Segmentation and reassembly (SAR) protocol is
  implemented in hardware in the board
• Incoming and outgoing channels operate
  simultaneously
• Congestion control is not needed as DTM uses
  circuit switching
• The Interface is based on shared memory in order
  to reduce the number of transfers over the host
  systems bus
• The interface uses interrupts for comm state
  changes with the processor
• The interface is designed to support fast
  connection establishment. Buffers are allocated
  to channels without involving CPU

                                                
   Fig 4. DTM Host Interface
32
DTM Host Interface

• A Low-Cost Interface to the DTM Network
• Proposed for equipment that doesnt need
  processing capability such as a video terminal.

• Hardware Aspects of the Prototype Interface
• The hardware implementation of the host interface
  was done between a DTM Network and SUN
  SPARCstation.

33
DTM Host Interface

• Performance Measurements
• Performance measurements were done using
  Sparcstation 2 and 10 and with memory buffers
  from 4KB to 1MB, in systems using different cache
  mechanisms
• Measurements show that the less System Bus access
  the better. The faster the system bus the better.
  Performance will also increase if no no update to
  the main memory is required when the buffer is
  flashed.

34
Conclusion

• With evolution of fiber optics, the bottleneck in
  the communication infrastructure is moving from
  capacity in the links to place in the network
  where processing and storage is needed, for
  example in switch points.
• The main argument is circuit-switched networks
  have a simple communication architectures with
  low requirements on processing of data elements
  that make them well-suited to handle high
  capacity in fibers.
• Convergence of communication, computing and media
  imposes new requirements on the services the
  network must provide.

35
Research Methodology Used

• Introduction
• Domain
• Problem purpose
• Expected Results
• General Background
• Framework in time and Technology
• Development and Experimentation
• Conclusions

36
What I have learnt through this reading

• Nothing that we say today is guaranteed to remain
  true for tomorrow. Anything we say could be valid
  within a certain timestamp even if the time stamp
  is too large for our life
• This study was done under some assumptions
• WDM precision diff not economically viable
• Cost of laser beamers and sensors were very high
  to make it viable.

37
Contribution to the Knowledge

• Knowledge so conceived is not a process that
  converges toward an ideal view it is an ever
  increasing ocean of alternatives, each of them
  forcing the others into greater articulation, all
  of them contributing, via this process of
  competition, to the development of our mental
  faculties
• Prof Naess

38
Tack!