Computer Networks Laboratory Electrical Engineering Department Technion, Haifa, Israel REAL TIME ET

1
     Computer Networks Laboratory
    Electrical Engineering Department         
Technion, Haifa, IsraelREAL TIME ETHERNET

• Presnting Ofir Blatt
• Amir Keren
• Supervisor Ilan Hazan

2
PROJECT GOAL

• Design, implementation and evaluation of a
  protocol that provides real-time performance
  guarantees to multimedia applications without
  requiring any modifications to existing
  half-duplex Ethernet (LAN) hardware.

3
PROJECT OBJECTIVES

• Examine the present Ethernet protocol (CSMA/CD)
  and requirements for real time network traffic.
• Design and implement a protocol that allows real
  time and non- real time network traffic.
• Simulate the new protocol and evaluate the
  simulation results.

4
Real time data

• Multimedia application incorporate continuous
  media data, such as video and audio, therefore a
  time constrain on network traffic is demanded.

• In order to preserve the characteristics of
  continuous media data, it is essential to
  guarantee resources (QoS) on these three
  domains network, I/O and processor.

5
IEEE 802.3 Ethernet protocol

• 1 persistent CSMA/CD
• Carrier Sense Multiple Access / Collision
  Detection.

• Binary exponential backoff

• T/R rate of 10Mbps (100M/1G/10G bps on fast
  Ethernet)

6
Ethernet and real time applications ?

• The probabilistic nature (backoff mechanism) of
  the protocol cannot provide bondable and
  deterministic access to the network.

• Ethernet packets dont have priorities,
  therefore it cannot distinguish between real
  time and non-real time data.

7
REAL TIME ETHERNET (RTE)
General description

• Each RTE station can transmit real time and non
  real time data.
• For non real time data RTE stations perform as
  ordinary Ethernet stations.
• Each RTE station has its own priority.

8

• Real time data is sent by RTE stations as fix
  sized frames periodically.

• Between periods, non real time data can be
  transmitted by all stations.

9
Real time packet
10
Real time chain structure
11
Constructing / joining a chain

• Collision between RTE stations can occur only
  when several stations try constructing or joining
  a chain simultaneously.
• Upon collision between RTE stations, each station
  will continue to transmit the head for a max time
  of N2? (where N is the stations priority) or
  until the station wins the channel.
• The station that folded, tries joining the chain
  immediately after the station that won the
  channel.

12
IDLE
????? ????? ?? ???????
????? ????? ?? ???????
???? ???? ??? ???
WAIT
?? ????? ?????/??? ????? ?????, ?? ????? ????
???? ??????? ????? ????? ????
??? ????? ?????/ ?? ????? ????? ??????? ?????? ???
?? ?????? ?? ????? ??????, ??????? ?? ????????
FIRST
CHAIN
13
Real time frame

• Both Dest Addr and Src Addr fields are the same.
• ID field - has the ID (priority) of the
  transmitting RTE station.
• Next field - has the ID of the next RTE station
  in the chain (0 - for non).
• Status field - Bit 0 - first packet in the
  chain.
• Bit 1 - last packet of
  the station.
• Bit 2 - last in chain
  (Next field 0).
• Other bits - equal to 0.

14
IMPLEMENTATION OF THE PROTOCOL

• Software (C/C)
• Matlab/Simulink
• Opnet
• HDL (Verilog/VHDL)

15
IMPLEMENTATION DESCRIPTION

• Verilog Modules descriptions

SYS (waveforms, seed)
NET (act,rt_state,transmitter,next,fic,finish)
STATION (idle,wait, first,chain)
STATION (idle,wait, first,chain)
STATION (idle,wait, first,chain)
STATION (idle,wait, first,chain)
16
Traffic management

• Non real time traffic
• Poissonic appearance.
• Always available.
• Real time traffic
• Exponential delay between files.
• Random file size, uniformly distributed.
• Constant packet size.

17
Simulation statistics

• For non real time packets
• Number of packets reached the buffer.
• Number of packets dropped due to buffer fullness.
• Number of packets that succeeded to transmit.
• Number of packets that dropped due to 16
  backoffs.
• Number of packets that remained in the buffer at
  the end of the simulation.
• Average size of packets.
• Average time which packet waits at the buffer.
• Average time at which a packet got to the head of
  the buffer until it was successfully transmitted.

18
For Real time packets

• Number of files reached to the stations.
• Number of packets successfully sent.
• Number of packets to be sent until end of file.
• Average number of RTE backoffs.
• Average time on constructing chain or joining to
  existing chain (due to backoffs) .
• Average jitter time.

19
Simulation outputs and results

• Only non-real-time stations.
• Packets always available.

20

• Only non-real-time stations.
• Poissonic appearance of packets.

21

• 1 real-time station.
• nrt Packets always available.

22

• Several real-time station.
• nrt Poissonic appearance of packets.

23
Simulation waveform
24
Summery and conclusions

• The implementation is a correct model of the new
  protocol.
• The protocol guarantees a constant bandwidth to
  all RTE stations as long as the upper bound of
  their number is kept.
• The new protocol coexist with the original
  Ethernet protocol.
• Real time traffic might harm other non real time
  traffic on the net, by taking a part of the
  channel bandwidth.
• The new protocol stands all the requirement for
  transmitting real time data over Ethernet, while
  allowing regular traffic on the net.

25
THE END