Lecture 2. Switching of physical circuits.

Slides:



Advertisements
Similar presentations
Optimization Problems in Optical Networks. Wavelength Division Multiplexing (WDM) Directed: Symmetric: Undirected: Optic Fiber.
Advertisements

Optical networks: Basics of WDM
Lecture 4. Topics covered in last lecture Multistage Switching (Clos Network) Architecture of Clos Network Routing in Clos Network Blocking Rearranging.
Dynamic Topology Optimization for Supercomputer Interconnection Networks Layer-1 (L1) switch –Dumb switch, Electronic “patch panel” –Establishes hard links.
Joining LANs - Bridges. Connecting LANs 4 Repeater –Operates at the Physical layer no decision making, processing signal boosting only 4 Bridges –operates.
Super computers Parallel Processing By: Lecturer \ Aisha Dawood.
1 Routing and Wavelength Assignment in Wavelength Routing Networks.
Management Science 461 Lecture 2b – Shortest Paths September 16, 2008.
1 Switching and Forwarding Bridges and Extended LANs.
DS - IV - TT - 1 HUMBOLDT-UNIVERSITÄT ZU BERLIN INSTITUT FÜR INFORMATIK DEPENDABLE SYSTEMS Vorlesung 4 Topological Testing Wintersemester 2000/2001 Leitung:
048866: Packet Switch Architectures Dr. Isaac Keslassy Electrical Engineering, Technion Scaling.
EE 4272Spring, 2003 Chapter 9: Circuit Switching Switching Networks Circuit-Switching Networks Circuit-Switching Concept  Space-Division Switching  Time-Division.
7. Fault Tolerance Through Dynamic or Standby Redundancy 7.6 Reconfiguration in Multiprocessors Focused on permanent and transient faults detection. Three.
CSE 291-a Interconnection Networks Lecture 7: February 7, 2007 Prof. Chung-Kuan Cheng CSE Dept, UC San Diego Winter 2007 Transcribed by Thomas Weng.
Color Aware Switch algorithm implementation The Computer Communication Lab (236340) Spring 2008.
Spring 2004 EE4272 The Need of Spanning Tree Algorithm.
1 K. Salah Module 4.3: Repeaters, Bridges, & Switches Repeater Hub NIC Bridges Switches VLANs GbE.
Introduction to Computer Networks 09/23 Presenter: Fatemah Panahi.
Interconnect Network Topologies
Interconnect Networks
ATM SWITCHING. SWITCHING A Switch is a network element that transfer packet from Input port to output port. A Switch is a network element that transfer.
Dynamic Interconnect Lecture 5. COEN Multistage Network--Omega Network Motivation: simulate crossbar network but with fewer links Components: –N.
EE384y EE384Y: Packet Switch Architectures Part II Scaling Crossbar Switches Nick McKeown Professor of Electrical Engineering and Computer Science,
1 Multicasting in a Class of Multicast-Capable WDM Networks From: Y. Wang and Y. Yang, Journal of Lightwave Technology, vol. 20, No. 3, Mar From:
Minimal Spanning Tree Problems in What is a minimal spanning tree An MST is a tree (set of edges) that connects all nodes in a graph, using.
1 Data Link Layer Lecture 23 Imran Ahmed University of Management & Technology.
Interconnect Networks Basics. Generic parallel/distributed system architecture On-chip interconnects (manycore processor) Off-chip interconnects (clusters.
Introducing a New Concept in Networking Fluid Networking S. Wood Nov Copyright 2006 Modern Systems Research.
Based on An Engineering Approach to Computer Networking/ Keshav
Multicast Communications
1 Lecture 15: NoC Innovations Today: power and performance innovations for NoCs.
Spring 2000CS 4611 Router Construction Outline Switched Fabrics IP Routers Extensible (Active) Routers.
© Sudhakar Yalamanchili, Georgia Institute of Technology (except as indicated) Deadlock: Part II - Recovery.
Lecture 20. Graphs and network models 1. Recap Binary search tree is a special binary tree which is designed to make the search of elements or keys in.
Assignment 1  Chapter 1:  Question 11  Question 13  Question 14  Question 33  Question 34  Chapter 2:  Question 6  Question 39  Chapter 3: 
Network layer (addendum) Slides adapted from material by Nick McKeown and Kevin Lai.
Computer Networks, Fifth Edition by Andrew Tanenbaum and David Wetherall, © Pearson Education-Prentice Hall, 2011 The Network Layer Chapter
INTERCONNECTION NETWORK
Yiting Xia, T. S. Eugene Ng Rice University
Chapter 3: Packet Switching (overview)
EE384Y: Packet Switch Architectures Scaling Crossbar Switches
Lecture 4 Sorting Networks
Interconnect Networks
Data Center Network Architectures
Networking Devices.
Dynamic connection system
ECE 565 High-Level Synthesis—An Introduction
Switching and High-Speed Networks
Refer example 2.4on page 64 ACA(Kai Hwang) And refer another ppt attached for static scheduling example.
Network Layer Goals: Overview:
Switching and Forwarding Bridges and Extended LANs
NTHU CS5421 Cloud Computing
Lecture 17: NoC Innovations
Chapter 5 The Network Layer.
Packet Switching (basics)
What’s “Inside” a Router?
Bridges and Extended LANs
SAT-Based Area Recovery in Technology Mapping
L12. Network optimization
COMPUTER NETWORKS CS610 Lecture-42 Hammad Khalid Khan.
High Performance Computing & Bioinformatics Part 2 Dr. Imad Mahgoub
EE 122: Lecture 7 Ion Stoica September 18, 2001.
CS 584 Lecture7 Assignment -- Due Now! Paper Review is due next week.
Project proposal: Questions to answer
Lecture 6 Dynamic Programming
Techniques and problems for
Connectors, Repeaters, Hubs, Bridges, Switches, Routers, NIC’s
Circuit Switch Design Principles
Design Principles of Scalable Switching Networks
Multiprocessors and Multi-computers
Presentation transcript:

Lecture 2. Switching of physical circuits

Crossbar Switch Concept Each egress port can select any ingress port as its source. Broadcast- a copy is taken by all egress ports. Multicast – a copy is taken by a set of egress ports.

Non-symmetric switch Ne – number of egress ports Ni – number of ingress ports In symmetric switch, Ne = Ni If Ne > Ni, then the crossbar is with speedup. Increase the number of paths that a set of signals can take. If Ne < Ni, then the crossbar is with slowdown. Decrease the number of paths that a set of signals can take.

Crossbar with speedup

Crossbar with slowdown

Broadcast by ingresses Broadcast is expensive and not practical.

Using fanout buffers Fanout tree amplifies a signal into N copies for each egress.

Cheaper Implementation

Single / Multi-Stage Switch Single-stage switch (one central node) has the best performance. However building arbitrarily large single-stage switch is impossible. We are forced to make multi-stage switches.

Multi-Stage Switch A 3-stage switch with 2 rows:

Clos Switch Structure Each Clos switch is represented by C(n,m,r).

Clos switch flow

Clos switch flow – cont.

Clos switch flow – cont

Clos switch with a speedup A search for available path through stage two is more likely to succeed. Also useful in case of link failure. In general, speedup = m/n

Blocking in Clos switch Blocking - there is no available route from free input to free output. Clos network can suffer from blocking. For example: (suppose two routes can not share a link) Strict non-blocking: can find a route from free input to free output without changing existing routes. Rearrangable non-blocking: can route any set of pairs (destroying existing state)

Clos network is strictly non-blocking (without rearrangement) iff m >= 2n-1. Proof: n-1 busy 1 n-1 busy … … … free free n-1 n … 2n-2 Number of switches at stage 2 >= 2n-1 Sometimes this switch arrangement is called “2n-1 switch” Speedup = (2n-1)/n

In practice, strictly non-blocking Clos is usually built with speedup = 2.

Clos network is rearrangable non-blocking iff m >= n Proof: the arrangement if all the services are active- 1 2 … n Number of switches At stage 2 >= n. Optimally, each building block of stage 1 has 2n ports, n ports to the outside and n ports to the stage 2, there are n building blocks at stage 2 and there are n building blocks on first and third level.

Algorithm to rearrange the paths Map the network to coloring edges on a graph:

Solving conflict in the graph: find open path Example: (b)

(c) (d)

Another example:

Multicast in clos networks Multicast in Clos is harder to implement than in Crossbar. Suppose A wants to send signal to both B and C. There are two options: Multiply signals at stage 1. Multiply signals at stage 2.

Enhanced Multicast in Clos Multicast branching: Best solution in terms of replication Harder to manage and configure the multicast trees

Single vs. Multi Stage There is no blocking in Single stage Single stage is less expensive No algorithm to find a path is need for Single stage. There is a physical limit to implement a Single stage above certain number of ports.