Review of Last Class 1. Signals and Spectrum 2. Transmission Media 3. Repeaters 4. Hubs 5. Packets versus Circuits 6. MAC layer 7. Token Ring Administrative 1. Pick up homework at break 2. Computer lab update 3. Projects Weblinks (reading) 1. http://www.networkuptime.com/faqs/ethernet/ 2. http://wwwhost.ots.utexas.edu/ethernet/enet-misc/capture/rich-seifert-on-capture-effect 3. http://www.networkmagazine.com/magazine/tutorial/internetworking/8906tut.htm 4. ftp://ftp.netcom.com/pub/se/seifert/TechRept14.pdf C. Traditonal LANs (cont.) Specific implementations of physical/MAC layer technologies. 1. Ethernet/802.3 A layer 1/2 bus technology which employes a shared medium access technology called carrier sense multiple access with collision detection (CSMA/CD). Typically deployed as a physical star wired, logical bus today (i.e. 10BASE-T), but also provides support for a physical bus wired, logical bus topology (10BASE5) or physical tree wired, logical bus topology (10BROAD36). a. Ethernet Mdia Designations i. 1BASE5 - UTP, 1 pair, 500 m max, "StarLAN", 1Mb/s ii. 10BASE5 - Thick coax, 500 m max, orig. Ethernet, 10Mb/s iii. 10BASE2 - Thin coax, 185m max, "Cheapernet", 10Mb/s iv. 10BROAD36 - 3 channels in each direction of private CATV 3.6 km max, 10Mb/s v. 10BASE-T - 2 pairs ofCat 3 or > UTP, 10Mb/s vi. 10BASE-F - generic designation for 10Mb/s fiber system 10BASE-FL - active hub, 2 km max 10BASE-FP - passive hub, 1 km max 10BASE-FB - synchronous active hub, 2 km max vii. 100BASE-T - generic 100Mb/s designation 100BASE-X - generic for 100BASE-T using 4B/5B encoding 100BASE-TX - 2pairs cat 5, 100 m max 100BASE-FX - 2 fibers, 2 km max 100BASE-T4 - 4 pair cat 3 or > UTP,100 m max 100BASE-T2 - 2 pair cat 3 or > UTP, 100 m max viii. 1000BASE-X generic designation for 1000Mb/s using 8B/10B 1000BASE-CX - 2 pair 150 ohm STP, 25 m max 1000BASE-SX - 2 fiber using shortwave laser optics 1000BASE-LX - 2 fiber using longwave laser optics 1000BASE-T - 4 pair cat 5 UTP, 100 m max b. CSMA/CD Carrier sense multiple access with collision detection can be described using this simple logic process: 1. If the medium is idle, transmit. 2. Else, continue to listen to the medium and when it is available, transmit. 3. If a collision is detected, jam and back-off. 4. After back-off (algorith, with some randomness), try to transmit again if the medium is idle. 5. Goto step 3 until max try counter is reached. Historically derived from ALOHA, packet radio networks. Improved upon access scheme by taking advantage of low propagation delay (versus frame transmission time). Collision: When two or more station transmission interfere with one another. Jam: Continue to transmit signal when a collision is detected to ensure that everyone knows there was a collision. Back-off: When a collision is detected, each transmitting station waits a period of time before attempting to transmit again. Limited amount of retries (16) before giving up. Let one slot time = 512 bit times (1) Upon first collision, randomly choose among {0,1} slot delay (2) Upon second collision, randomly choose among {0,1,2,3} slot delay (3) Up to a maximum of 16 transmission attempts with a range of delay from {0 to 1024} slot times. 0 <= r < 2^k where k = MIN(n,10) where n is the n-th transmission attempt c. Ethernet/802.3 MAC frame format 1. Preamble (7) 2. Starting frame delimiter (1) 3. Destination Address (6) 4. Source Address (6) 5. Length/Type (2) 6. Data (46 to 1500) 7. Frame Check Sequence (4) d. Collision Domain Using CSMA/CD, there is a relationship between the minimum frame size and the maximum round-trip propagation delay of the Ethernet network. Ex: Two stations colliding. One station is at the far end and it's frame is too short, so that it finishes transmission before it detects a collision. ...or the network is too long before the collision reaches it. Thus, the minimum length of the frame must be longer than the maximum round-trip propagation time of the network, plus an allowance for jam, sync delays, etc. So, the trade-off is how long do we make the minimum frame size (potentially wasting bandwidth with padding) and how short do we make the network (limiting reach). The minimum frame is 512 bits (64 bytes) not including preamble and starting frame delimiter. This means we need at least 46 bytes in the data field. We can pad if necessary. IP/TCP will use 40 bytes minimumally. 512 bits is 51.2 microseconds at 10Mb/s. This allows for up to 2 or 3 kilometers for a Ethernet network. Note that distance then would decrease as the speed increases. 100Mb/s systems minimum frame of 512 bits takes only 5.12 microseconds. Standards for distance limitations was reduced to 200 meters (using twisted pair). This required no change in the algorithm. If this same solution were to be adopted for Gigabit Ethernet, the maximum distance for the network would be only 20 meters! Solution involved: 1. keep 512 bit frame minimum as seen by software (drivers) 2. modify MAC to use a 'carrier extension' 3. support frame bursting In practice, this is not used, we do full-duplex. Stay tuned. d. Auto-negotiation/configuration Auto-negotiation allows the station and hub/switch port to automatically determine each other's capabilities and configure themselves to the "highest" common denominator. - Half/Full Duplex - Speed (i.e. 10/100/1000) - Flow control Used for xBASE-T based systems only. e. Half-duplex/full-duplex Half-duplex mode is the standard operating mode of Ethernet CSMA/CD enabled. In full duplex mode, a point-to-point connection is needed so that CSMA/CD is turned off. Both ends can transmit and receive simultaneously. No longer a shared medium. Requires a "switch". Stay tuned. We've lost most of our MAC layer! f. Inter-frame gap -96 bit times for "housekeeping" chores g. Ethernet maintenance i. Collisions ii. Late collisions iii. short frames (runts) D. LAN Bridging/Switching Book: Interconnections, Second Edition: Bridges, Routers Switches and Internetworking Protocols, Radia Perlman, ISBN: 0201634481 CIO of a prominent business consulting firm in the 1980's once said: "Why would anyone want to connect two LANs together?" Discussion on why. Note: A bridge is a switch and a switch is a bridge. Bridge refers to older hardware with limited capabilities and performance. Switching is mostly a marketing term for the same function, now sped up through hardware enhancements and lower manufacture costs. We will use the terms bridge and switch interchangeably in this class. A traditional layer 2 switch connects 1 or more common data link networks together. Ex: Ethernet to Ethernet. Token Ring to Token Ring. Not Ethernet to Token Ring. Characteristics of a Brige/Switch o Two or more network interfaces, connecting to two or more data link networks. o Receives and inspects all frames on it's data link network interfaces regardless of the destination MAC address. o Makes a "bridging/switching" decision based on information contained in the MAC frame fields and typically the associated information within it's own internal "bridge/switch" tables. o Potentially forwards traffic from one data link network to another based on the decision and inforamation previously described. Typical functions performed by a bridge/switch o Separates LAN workgroups for traffic efficiency and reliability. - utilization - collision domain - redundancy o Performs MAC based filtering for security or performances reasons. Note: Except for special cases in source route bridging, bridges do not modify the frames they handle. Except in source route bridging, end stations are oblivious to the bridge's existence. A. Transparent Bridging 1. Address Learning - bridge builds MAC address tables - unknown destination addresses "flooded" - new stations learned and put into table 2. Frame Forwarding 3. Spanning Tree - bridges communicate with each other to determine topology and create loop-free paths - designate a root bridge - build tree so that only one path from all data link networks to the root 4. BPDU - configuration - topology change B. Source Route Bridging IBM Token Ring and the IEEE 802.5 standard define an alternative procedure for moving frames between data link networks. The idea is to put the intelligence at the end stations, where bridging "routes" are computed. source route indicators: - null: stations on same data link network - non-broadcast: frame includes RIF as necessary - all route broadcast: search all LANs, repsonse back is direct - single route broadcast: direct to destination, arb back 1. Route Discovery 2. Frame format - routing information field - segment numbers (ring id's) - bridge id's C. Switching Improvements to Layer 2 The primary contribution of switches to the LAN market are due to advances in hardware (ASICs). These advances have brought higher port densities, low latency data link network connectivity, and VLANs to market among other technologies. You may think of LAN switches, as fast, multi-port bridges with some "new" features in how they manage LAN traffic. 1. Store and forward 2. Cut-through 3. VLANs Group Scenarios