![Chapter 1
Autonomous & Dynamic
Forwarding Tables
• Layer 2 Control
– Solution
• Spanning Tree Protocol (STP) [IEEE 802.1D]
– All switches in the topology compute a spanning tree
» “One path to every leaf (MAC address) on the tree”
Page 28
Other protocols have superseded STP
• Shortest Path Bridging (SPB)
• Transparent Interconnection of Lots of
Links (TRILL)](https://image.slidesharecdn.com/sdnchapter1-introduction-250108161752-9ae3be6b/85/software-defined-networks-Introduction-pptx-28-320.jpg)
software defined networks Introduction.pptx
- 1. Chapter 1 1 Software Defined Networks Chapter 1 Introduction
- 2. Chapter 1 These slides have been prepared to accompany the book “Software Defined Networks – A Comprehensive Approach” Dr. Paul Goransson, Chuck Black, and Timothy Culver. The authors and publisher Morgan Kaufman allow lecturers and public and private universities the right to modify these slides for their own use when the book is being used for a class. Since the SDN marketplace is changing rapidly, the accompanying PowerPoints and related classroom material will be updated every few months Page 2
- 3. Chapter 1 SDN • What is SDN? – Fundamentally novel way to program switches in modern data networks. – Designed from the onset to make fine grained traffic forwarding • Evolution versus Revolution • Why SDN? – To understand SDN and its value proposition, chapter 1 provides a background on the current state of the art Internet switches Page 3
- 4. Chapter 1 Basic Packet Switching Terminology • An old boss once told me “The more network acronyms that you know, the higher you will get paid!” • The appendix has a list of all acronyms used in the book. • Key ones – WAN – LAN – MAN – WLAN Page 4
- 5. Basic Packet Switching Terminology • Physical layer (a.k.a. Layer 1) – Consists of the basic hardware transmission technology to move bits of data on a network. Chapter 1 Page 5
- 6. Basic Packet Switching Terminology • Data Link Layer (a.k.a. Layer 2) – Transfers data from one device to another on a single network segment Chapter 1 Page 6 LAN Segment LAN Segment LAN Segment Repeater Repeater For this book, LAN Segments connected by repeaters are treated as a single LAN segment
- 7. Basic Packet Switching Terminology • Data Link Layer (a.k.a. Layer 2) – MAC (Media Access Control) is part of the Data Link Layer Chapter 1 Page 7
- 8. Basic Packet Switching Terminology • Network Layer (a.k.a. Layer 3) – Port • Connection to single communications medium • Interface and port are terms that are interchangeable in the book. – Virtual Interfaces Chapter 1 Page 8
- 9. Basic Packet Switching Terminology • More terms – Packet – Frame – Switch • Receives information on one port and retransmits on another port – Circuit Switch Chapter 1 Page 9 Packet Frame Packet Layer 2 Layer 3 Circuit Switch Call Setup Connection Maintained Port
- 10. Basic Packet Switching Terminology • Packet Switch – Connection oriented – context information retained to forward to same address. – Connectionless – data transmits to destination independently. Chapter 1 Page 10 Packet Switch Packet 2 Packet 1 Packet 2 Packet 1
- 11. Basic Packet Switching Terminology Chapter 1 Page 11 Router • Router • Flood a packet – broadcast on all ports • Broadcast a packet – same as flood • Line Rate – bandwidth on medium connected to port WAN 192.168.1.23 192.168.1.25 192.168.1.33 Subnet 192.168.xxx.xxx
- 12. Chapter 1 Historical Background • 1st half of 20th century – Telephone networks – circuit switched – Rand Corporation Study • Paul Baran – Nuke a major switching Center and communications out for a large portion of the country Page 12
- 13. Chapter 1 Historical Background • Solution – Baran had the idea to send data in packets that traveled autonomously through the network! • If switching center destroyed, packet would be resent. – ARPANET – 1969 • Built the original Internet – Vint Cerf (father of the Internet) – co-patent holder of the TCP/IP protocol – Now you know where cerfing the Internet came from! Page 13
- 14. Chapter 1 Modern Data Center • WWW leads to – Large data centers with huge numbers of servers and even greater VMs Page 14 In 2014, 80% of traffic is East / West Here
- 15. Chapter 1 Modern Data Center • Protocols designed to provide robustness over geographic area not appropriate for the huge amount of traffic in the DC – Routers spend 30% of CPU cycles in rediscovery and recalculating routes – East / West traffic doesn’t benefit from overhead of protocols. – SDN was designed to handle the network of the modern data center fundamental shift from traditional Internet switching Page 15
- 16. Chapter 1 Traditional Switch Architecture • Traditional Internet switch from an architecture perspective – Data Plane – Control Plane – Management Plane Page 16
- 17. Chapter 1 Traditional Switch Architecture Page 17
- 18. Chapter 1 Traditional Switch Architecture • Data Plane – Vast majority of packets only touch data plan – Contains • Reception & Transmission Ports • Forwarding Table • What does the Data Plane do? – Packet buffering – Packet scheduling – Header modification and forwarding • What happens if the packet information is not in the forwarding table? – Data Plane communicates vertically to Control Plane Page 18
- 19. Chapter 1 Traditional Switch Architecture • Control Plane – Principal role - Keep forwarding table up-to-date – Other roles • Process control protocols - Control protocols collectively manage the topology of the network. Page 19 Control Protocols may be complex enough to require micro processors & software in the control plane
- 20. Chapter 1 Traditional Switch Architecture • Management Plane – Network administrators configure and monitor the switch through this plane – Interfaces vertically to collect or update information in other planes – Typically a NMS (network management system) communicates to the plane in the switch. • Check out – openNMS – Cisco NMS Best Practices Page 20
- 21. Chapter 1 Traditional Switch Architecture • Software-Based Routing and Bridging – Battle of the 80’s – Early Connectionless Routers • Unix box with a lookup table to determine where to forward the pack (layer 3 router) • Control packets sent to control processes versus being processed in routing table. Page 21 Connection Oriented Packet Switching vs. Connectionless Router
- 22. Chapter 1 Traditional Switch Architecture • Software-Based Routing and Bridging – Bridges • Layer 2 • Created a topology of interconnected LAN segments Page 22 Token Ring FDDI Ring Bridge
- 23. Chapter 1 Traditional Switch Architecture • Software-Based Routing and Bridging – Ethernet interface speed increased through the 90’s. – Hardware solutions leveraged to help routers and bridges keep up with increasing speed – Eventually software could not keep up with header inspection and routing table lookups Page 23
- 24. Chapter 1 Traditional Switch Architecture • Hardware Lookup of Forwarding Tables – 1st major hardware acceleration in packet switching • Application-specific integrated circuits (ASICs) – Content Addressable Memory (CAM) High speed lookups! – Debut in Ethernet Switches • ASICs – In the beginning, layer 2 lookup (MAC addresses) • Today, the terms Layer 2 and Layer 3 switch are just referred to as switch Page 24 Mid-90’s
- 25. Chapter 1 Traditional Switch Architecture • Generically Programmable Forwarding Rules – Early routers – limited packet header field mods. – Switch features grew over time • Multi-cast, VLANs, MPLS, etc. • Pushing programmable rules to the hardware allows complex manipulation while maintaining line rates Page 25 Programmable hardware gives life to the concept of SDN
- 26. Chapter 1 Traditional Switch Architecture • Generically Programmable Forwarding Rules Page 26
- 27. Chapter 1 Autonomous & Dynamic Forwarding Tables • Layer 2 Control – Old way • Forwarding table indexed by destination MAC address • Switch learns sender’s MAC address/location and updates the forwarding table • Unknown address – – Frame goes out to all interfaces – Problem – Bridges interconnected into networks of bridges this will create a infinite loop » Problem is really a problem because Layer 2 does not have a Layer 3 TTL capability Page 27
- 28. Chapter 1 Autonomous & Dynamic Forwarding Tables • Layer 2 Control – Solution • Spanning Tree Protocol (STP) [IEEE 802.1D] – All switches in the topology compute a spanning tree » “One path to every leaf (MAC address) on the tree” Page 28 Other protocols have superseded STP • Shortest Path Bridging (SPB) • Transparent Interconnection of Lots of Links (TRILL)
- 29. Chapter 1 Autonomous & Dynamic Forwarding Tables • Layer 3 Control – Fundamental task of router - Decide • Forward packet out one of the interfaces or • Perform exception processing on packet – Routing table • Contains Layer 3 networks and sometimes Layer 3 host addresses • Built through a number of routing protocols Page 29
- 30. Chapter 1 Autonomous & Dynamic Forwarding Tables • Layer 3 Control – Interior Gateway Protocols (optimize routing in AS) • Routing Information Protocol (RIP) • Open Shortest Path First (OSPF) • Intermediate System to Intermediate System (IS-IS) – Exterior Gateway Protocols • Border Gateway Protocol (BGP) Page 30
- 31. Chapter 1 Autonomous & Dynamic Forwarding Tables • Protocol Soup or Switch’s Brew – Control protocols on previous slide are a small list of protocols that need to be implemented in a modern router • LDP for MPLS labels • IGMP • MDSP • PIM Page 31
- 32. Chapter 1 Autonomous & Dynamic Forwarding Tables • Protocol Soup or Switch’s Brew – Past control protocols – – Autonomous switches to respond to network changes – Modern Data Center (24x7) • Topology Stable • Topology Strictly Controlled • Failures Quickly Addressed due to Global Topology Knowledge Page 32
- 33. Chapter 1 Autonomous & Dynamic Forwarding Tables • Protocol Soup or Switch’s Brew – Centralized controller • Quickly re-provision forwarding tables. • Simple approach - driver for SDN. • Addresses instability in the control plane Page 33
- 34. Chapter 1 Autonomous & Dynamic Forwarding Tables • Protocol Soup or Switch’s Brew – Packet Forwarding IQ • Policy Based Routing – Different types of traffic over different routes – Supports centralized approach of SDN • Program appliances that have relied on deep packet inspection into the SDN switch – Paradigm shift • SDN focused on defining forwarding tables to act on flows versus mapping destination address to an output port. Page 34
- 35. Chapter 1 Autonomous & Dynamic Forwarding Tables • Open Source and Technological Shifts – Open source community to develop much of the SDN control plane software – Example Sources • Linux • OpenSSL – Early examples of collaboration on SDN • University environment Page 35
Editor's Notes
- #4: A wide area network (WAN) is a network that covers a broad geographical area, usually larger than a single metropolitan area. A local area network (LAN) is a network that covers a limited geographical area, usually not more than a few thousand square meters in area. A metropolitan area network (MAN) is a network that fills the gap between LANs and WANs. This term came into use because LANs and WANs were originally distinguished not only by their geographical areas of coverage but also by the transmission technologies and speeds that they used. With the advent of technologies resembling LANs in terms of speed and access control, but with the capability of serving a large portion of a city, the term MAN came into use to distinguish these networks as a new entity distinct from large LANs and small WANs. A wireless local area network (WLAN) is a LAN in which the transmission medium is air. The typical maximum distance between any two devices in a wireless network is on the order of 50 meters.
- #6: For our definition of data link layer, we consider multiple segments linked by repeaters as a single LAN segment The link layer includes: (1) mechanisms to detect sequencing errors or bit errors that may occur during transmission, (2) some mechanism of flow control between the sender and receiver across that network segment, and (3) a multiplexing ability that allows multiple network protocols to use the same communications medium. These three functions are considered part of the logical link control (LLC) component of the data link layer.
- #7: For our definition of data link layer, we consider multiple segments linked by repeaters as a single LAN segment The link layer includes: (1) mechanisms to detect sequencing errors or bit errors that may occur during transmission, (2) some mechanism of flow control between the sender and receiver across that network segment, and (3) a multiplexing ability that allows multiple network protocols to use the same communications medium. These three functions are considered part of the logical link control (LLC) component of the data link layer.
- #8: A frame is the unit of data transferred over a layer two network. A packet is the unit of data transferred over a layer three network.
- #9: A frame is the unit of data transferred over a layer two network. A packet is the unit of data transferred over a layer three network.
- #13: In the early days of the Internet, common thinking was that the connection oriented protocols like X.25 and ATM would be the primary mechanisms of high speed transfer. As the Internet evolved, the flexibility of the protocol allowed it to handle very high bandwidth once thought only possible with cell switching technology like ATM.
- #14: Example from book Data centers are being built now that will accommodate over 120,000 physical servers [8]. State-of-the-art physical servers can conceivably host 20 virtual machines (VMs) per physical server. This means that the internal network in such a data center would interconnect 2,400,000 hosts
- #17: The vast majority of packets handled by the switch are only touched by the data plane. The data plane consists of the various ports that are used for the reception and transmission of packets and a forwarding table with its associated logic. The data plane assumes responsibility for packet buffering, packet scheduling, header modification and forwarding. If an arriving data packet’s header information is found in the forwarding table it may be subject to some header field modifications and then will be forwarded without any intervention of the other two planes. Not all packets can be handled in that way, sometimes simply because their information is not yet entered into the table, or because they belong to a control protocol that must be processed by the control plane.
- #18: If an arriving data packet’s header information is found in the forwarding table it may be subject to some header field modifications and then will be forwarded without any intervention of the other two planes. Not all packets can be handled in that way, sometimes simply because their information is not yet entered into the table, or because they belong to a control protocol that must be processed by the control plane.
- #19: Keep forwarding table up to date on Data plane so that it can handle as much traffic as possible. Control protocols may affect the forwarding table, depending on the configuration and type of switch. These control protocols are jointly responsible for managing the active topology of the network.
- #20: Emphasize four aspects of NMS (Fault, Config, Performance, and Accounting
- #22: Bridges create a bridged LAN, which is a topology of interconnected LAN segments. Because multiple competing layer two technologies were prevalent, including Token Ring, Fiber Distributed Data Interface (FDDI).
- #24: ASICs could not handle layer 3 lookups in the beginning. The most significant bit lookup of layer 3 made it impossible. As the speed increased, layer three lookups were possible blurring the difference between a router and switch.
- #26: In the figure we see that the packet may transit the packet receive, ingress filter, packet translation, egress filter, and packet transmit functions or be consumed by the switch OS
- #27: Unknown address – Frame is dropped on all except for the one where the destination is located.
- #28: Unknown address – Frame is dropped on all except for the one where the destination is located. STP is still prevalent in the network today.
- #30: RIP – each router in the RIP-controlled routing domain periodically broadcasts its entire routing table on all of its interfaces. This includes hop count from the broadcasting router to he reachable network. This permeates the entire routing domain (autonomous system – AS). This is a distance vector protocol. OSPF – Link state protocol – maintains a complete and current view of the state of each link in the AS. Uses cost of edges Dijkstra’s algorithm. OSPF was a an IETF project ISIS – Link state protocol (origins in OSI) OSPF used in large enterprise networks … ISIS used in large service providers BGP – used by large Internet companies who have a large number of interconnected AS’s. Learns about networks that are reachable via peer edge routers Note that some private networks may need BGP because the multiple AS’s can’t grow any larger
- #31: 30% of router CPU is used to handle control plane activity.
- #32: 30% of router CPU is used to handle control plane activity.
- #33: Breaking the control plane out to be a separate unit.
- #34: Firewalls, load balancers, intrusion detection systems are examples of what can be programmed into the switch. The power and capabilities of the network can grow exponentially with SDN Note SDN switching will affect layer 2 and 3, not the application layer.