Infrastructure Components: Hub & Repeater. Network Infrastructure. Switch: Realization. Infrastructure Components: Switch

Transcription

1 Network Infrastructure or building computer networks more complex than e.g. a short bus, some additional components are needed. They can be arranged hierarchically regarding their functionality: Repeater Physically increases the range of a local area network Hub onnects several computers or local area networks of the same type (to a broadcast network) ridge onnects several local area networks (possibly of different types) to a large LN Switch Like a hub, but without broadcast Router onnects several LNs with the same network protocol over large distances Page 1 Infrastructure omponents: Hub & Repeater Transmission of data on the physical layer Reception and refreshment of the signal, i.e. the signals received on one port are newly produced on the other(s) Increase of the network range Stations cannot send and receive at the same time One shared channel (roadcast) Low security, because all stations can monitor the whole traffic Low costs Hub Hub: one to all Repeater: Linking of 2 networks Segment 1 Segment 2 Layer 1 Repeater Page 2 Infrastructure omponents: Switch Switch: Realization Like a hub, but: Point-to-point communication no broadcast Switch learns the addresses of the connected computers Stations can send and receive at the same time No carrier control necessary uffer for each individual station/each port Switch Higher costs Layer 3-Switch : also has functionalities of level 3, i.e. it can e.g. take over the routing. Layer 4-Switch : looks up additionally in the TPheader, can therefore be used e.g. for load balancing. Layer 2/3/4 Mostly used: buffered crossbar or each input port, provide buffers for the output ports t any time, only one input port can be connected to an output line dditional speedup possible with small buffers at each cross-point With a buffered switch, each station has its own channel, influences from other stations are not longer given No more broadcast network, no collisions between transmissions of different stations Page 3 Page 4

2 Switches and ata orwarding To ease the use of switches, they have to learn on which port a host is connected Each switch maintains a forwarding table with entries <M address, port, age> M address: physical address of a host port: port number at switch where the host is connected to age: aging time of entry ssume a M frame arrives on port x: Switches - ddress Learning atabase entries are set automatically with a simple heuristic: the source address of a frame that arrives on a port tells which hosts are reachable from this port lgorithm: or each frame received, the switch stores the source address in the forwarding database together with the port where the frame was received ll entries are deleted after some time orward the frame on the appropriate port Is M address of destination in forwarding database for ports,, or? ound and x? Not found? ound and = x? Ignore frame roadcast the frame, i.e., send the frame on all ports except port x. Src=x, est=y Src=x, est=y Src=y, Src=x, est=x est=y Port 1 Port 2 Port 3 x is at Port 3 y is at Port 4 Port 4 Port 5 Port 6 Src=y, Src=x, est=x est=y Src=x, est=y Src=x, est=y Page 5 Page 6 LNs: us LNs: Star Till now: cables, repeater, hubs, switches what networks could we construct out of it? Terminating resistor Ω us roadcast Network: if station intends to send data to station, the message reaches all connected stations. Only station processes the data, all other stations are ignoring it. Passive coupling of stations - Restriction of the extension and number of stations to connected (but: use repeaters to connect two busses and by this increase the network extent) + Simple, cheap, easy to connect new stations + The breakdown of a station does not influence the rest of the network Ω Example: Ethernet Page 7 Example: ast Ethernet Star esignated computer as central station: a message of station is forwarded to station via the central station roadcast network (Hub) or point-topoint connections (Switch) Expensive central station Vulnerability through central station (Redundancy possible) + N connections for N stations + Easy connection of new stations Page 8

3 LNs: Tree LNs: Ring ranch 1 ranch 2 Repeater Router ackbone Tree Topology: onnection of several busses or stars ranching elements can be active (Router) or passive (Repeater) + ridging of large distances + daptation to given geographical structure Example: Token Ring, I Ring roadcast Network hain of point-to-point connections ctive stations: messages are regenerated by the stations (Repeater) reakdown of the whole network in case of failure of one single station or connection + Large extent possible + Easy connection of new stations +Only N connections for N stations Variant: bidirectional ring stations are connected by two opposed rings + Minimization of the cable length possible Page 9 Page 10 LNs: Meshed Networks ully Meshed Network Point-to-Point connections between all stations N( N 1) or N stations, 2 connections are needed onnecting a new station is a costly process + Redundant paths + Maximal connection availability through routing integration LNs: Examples Ethernet (IEEE 802.3, 10 Mit/s) - originally the standard network - available in an immense number of variants Token Ring (IEEE 802.5, 4/16/100 Mit/s) - for a long time the Ethernet competitor - extended to I (iber istributed ata Interface) ast Ethernet (IEEE 802.3u, 100 Mit/s) - at the moment the most widely spread network - extension of Ethernet for small distances Partly meshed network: cheaper, but routing, flow control, and congestion control become necessary (Wide rea Networks) Gigabit Ethernet (IEEE 802.3z, 1,000 Mit/s) - very popular at the moment; 10 Git/s are already in the planning phase at the moment Page 11 Page 12

4 Infrastructure omponents: ridges Infrastructure omponents: ridges With bridges, several LNs are connected on the link layer possibly LNs of different types i.e. having different header formats Layer 2 1) Transparent bridges (e.g. for SM and token bus networks) Major tasks: ppropriate forwarding of the data daptation to different LN types Reduction of the traffic in a LN segment, i.e. packets which are sent from to are not forwarded by the bridge to LN 2. Thus, station can communicate with E in parallel. Increases physical length of a network Increased reliability through demarcation of the LN segments 1 LN LN 2 1 E 1 LN LN 2 LN 3 1 haracteristics: oupling of LNs is transparent for the stations (i.e. not visible) Hash tables contain the destination addresses Source and destination LN are identical frame is rejected by bridge (e.g. 1 in case of a transmission from to ) Source and destination LN are different pass on frames (e.g. in case of a transmission from to E) estination LN unknown pass on frame ² E Page 13 Page 14 Loops Spanning Tree ridges onsider two LNs that are connected by two bridges. ssume host n is transmitting a frame with unknown destination. ridges and flood the frame to LN 2. ridge sees on LN 2 (with unknown destination), and copies the frame back to LN 1 ridge does the same. The copying continues Solution: Spanning Tree lgorithm ridge LN 2 LN 1 ridge Preventing loops: compute a spanning tree from all bridges connected Spanning Tree lgorithm: etermine one root bridge (the bridge with the smallest I) etermine a designated bridge for each LN (the bridge which is nearest to the root bridge) etermine root ports (port for the best path to root bridge considering costs for using a path, e.g. number of hops) LN 2 ridge 3 ridge 1 LN 5 ridge 5 LN 1 ridge 2 d ridge 4 host n LN 3 LN 4 Page 15 Page 16

5 Spanning Tree lgorithm Spanning Tree lgorithm: Example t the beginning, all bridges assume to be root bridge and send out a packet containing their own I and current costs (initialized with zero) over all of their ports: root I costs bridge I port I e.g. for station on port P 1 : 0 P 1 bridge receiving such a packet checks the root I and compares with its own one. Root I and costs are updated for received packets with smaller I in the root bridge field, and forwarded. Updating the costs is made by adding the own costs for the station from which the packet was received to the current costs value. When the (updated) packets of all bridges have passed all other bridges, all bridges have agreed on the root bridge. The received packets containing the smallest costs value to the root bridge determine the designated bridge for a LN and designated ports for the bridges to send out data. Page 17 Network: LN I=27 8 LN 4 Spanning Tree: LN 1 2 I=27 LN I=93 20 LN I= I=93 LN 3 5 I= I=18 4 I=3 10 LN 5 3 I=18 4 I=3 LN 5 7 ports LN 2 I: bridge I : designated port LN 2 designated bridge for LN 2 root bridge Page 18 Infrastructure omponents: ridges Networks 2) Source Routing ridges (e.g. for ring networks) Ethernet 1 I ² Ethernet haracteristics: Sources must know (or learn), in which network segment the receivers are located Large expenditure for determining the optimal route, e.g. via using a Spanning Tree algorithms or sending out Route iscovery rames using broadcast ll LNs and ridges on the path must be addressed explicitly onnection-oriented, without transparency for the hosts Layer 1 and 2: How to physically transport data reliably from one computer to a neighbored one? Layer 1 defines transmission medium and bit representation on this medium Layer 1 additionally specifies transmission mode, data rate, pin usage of connectors, Layer 2 protects against transmissions errors (mostly R) and receiver overload (flow control, sliding window) Layer 2 also defines medium access coordination for broadcast networks oth layers together define a network topology and how to transfer data from one computer to a directly connected one (maybe over a hub/switch) on that reason both are implemented in one piece of software: the network interface card driver. ridges in principle allow to connect lots of LNs over long distances is that the Internet? Page 19 Page 20

6 Infrastructure omponents: Router What are the limitations of bridges? Even though bridges are suitable to connect computers in several networks, there are also some disadvantages, e.g.: ridges can support only some thousand stations, which especially has the reason that addresses are used which do not have any geographical reference. LNs coupled with bridges already form a large LN, although a separation often would be desirable (e.g. regarding administration or errors). ridges pass broadcast frames on to all attached LNs. This can result in roadcast Storms. ridges do not communicate with hosts, i.e. they do not hand over information about overload situations or reasons for rejected frames. Router overcome these weaknesses Infrastructure omponents: Router Prinicipal task of routers : Incoming packets are being forwarded on the best path possible to the destination on the basis of a global address In principle no restriction concerning the number of hosts (hierarchical addressing) Local administration of the networks (ends at the router), irewalls are possible roadcasts are not let through by the routers, Multicast depending on the router ommunication between host and router improves performance LN 1 R R R Network 1 Network 2 LN 2 Layer 3 More details to routers: chapter 3 Page 21 Page 22