[ Pobierz całość w formacie PDF ]
.If a station on VLAN2 broadcasts a packet, the switch broadcasts it onlyon ports in VLAN2 (in this case, only port a).However, if a station on VLAN3 broadcasts apacket, the switch must broadcast it on all ports in VLAN3 (in Figure 5.12, ports a and c).Ifenough stations move around, broadcast traffic (mainly ARP packets) and unknowndestination packets wind up having to be transmitted on all ports.You might as well havesimply assumed the entire network was all one big LAN.5.3.6 Dynamic VLAN Binding, Switch-SwitchNow let's look at the case when a switch can be attached to another switch (see Figure5.13).Figure 5.13.VLAN tagging needed on interswitch portOn the link between the two switches, packets can belong in either VLAN1 or VLAN2.SoIEEE standardized a scheme for adding additional information, known as a VLAN tag , to apacket so that switches can know which VLAN a packet is intended for.A station would beconfused if it received the packet with a VLAN tag, so the switches must be configured toknow which ports contain switches and which ports contain stations.A switch removes theVLAN tag from the packet before forwarding the packet onto a non-switch-neighbor port.The VLAN tag is a 2-byte quantity containing 3 bits for priority, 12 bits for a VLAN ID, and 1 bitindicating whether the addresses are in canonical format.You define its presence by usingthe Ethernet type 81-00.For example, on 802.3 there might be a packet that looks like thatshown in Figures 5.14 and 5.15.Figure 5.14.Ethernet packet without VLAN tag Figure 5.15.Same packet with VLAN tagAn Ethernet type is used even if the original packet was in 802.3 format.For example, thepacket in 802.3 format shown in Figures 5.16 and 5.17 is converted to one with a VLAN tagby adding the same 4 bytes (81-00 for Ethertype and 2-byte VLAN tag) as with an Ethernetpacket.Figure 5.16.802.3 packet without VLAN tagFigure 5.17.Same packet with VLAN tagOn LANs other than 802.3 it is not possible to use Ethernet format, so SNAP encoding isused to insert the VLAN tag.An example is shown in Figures 5.18 and 5.19.Figure 5.18.Packet without VLAN tagFigure 5.19.Same packet with VLAN tagNotice that in this format there is 10 bytes of overhead for the VLAN tag.If the committee hadwanted to reserve one of the precious SAP values for VLANs, it could have done so muchmore compactly.For example, it could have reserved a SAP value say, 97 and then ifdsap = 97 the VLAN tag would immediately follow.Then there would be only 3 bytes ofoverhead for the VLAN tag.Or if that would be considered unsavory (because the taggedpacket doesn't have a proper ssap and ctl field), the committee could set dsap and ssap to 97and leave the ctl field, requiring only 5 bytes of overhead for the VLAN tag.The overhead forthe tag is ignored in terms of maximum legal packet size.If the packet without the VLAN tagis of legal size, it's still considered of legal size with the tag inserted.A switch can learn all the VLANs available on the port connecting it to another switch basedon the VLAN tags in received packets on that port.Because stations are unlikely tounderstand the VLAN tag, the switch must be carefully configured so that it never transmits apacket with a VLAN tag on a port on which a station might reside.Some switch vendors have a proprietary mechanism for having one switch tell another switchwhich VLANs are reachable on the switch-switch port.In that way, instead of learning theVLANs available through the switch-switch port based on received packets, the switch learns it through explicit protocol messages.But because there is no standard (as of the writing ofthis book) for a switch to tell its neighbor switch which VLANs it is connected to, the twoswitches would have to be from the same vendor.Homework1.Suppose A, B, C, D, and E are attached to a layer 2 switch.Suppose A and B areattached with 10 Mb/sec ports, whereas C, D, and E are attached with 100 Mb/secports.Furthermore, assume that the switch has unlimited bandwidth, although limitedbuffering, so that it applies backpressure or flow control as needed to prevent itsbuffers from overflowing.Ignore reverse traffic; that is, when I say, "A is sending to B"assume a constant stream of traffic from A toward B.And assume that links are full-duplex.What is the total aggregate bandwidth in the following cases, assuming thateach station sends as quickly as it can?a.A is sending to B, C is sending to D.b.A is sending to B, C is sending to B.c.A is sending to C, B is sending to D.d.C is sending to A, D is sending to B.e.C is sending to D, E is sending to C.2.Why are collisions (as with layer 1 switches) not an issue with a layer 2 switch that isdoing cut-through forwarding?3.Why is it possible to have ports with differing speeds on a layer 2 switch but not on alayer 1 switch?4.How would you implement cut-through forwarding on a switch if ports are differentspeeds? What are the issues when forwarding from slow to fast links and from fast toslow links?5.Why would it be difficult for layer 1 switches (repeaters) to run the spanning treealgorithm?6.With a topology of multiple switches and multiple VLANs, many vendors choose torun separate instances of the spanning tree algorithm for each VLAN.In other words,each VLAN constructs its own spanning tree.Some links might belong to multipleVLANs and therefore belong to multiple spanning trees.What are the advantages ofrunning a different instance of the spanning tree for each VLAN instead of running asingle spanning tree?Chapter 6
[ Pobierz całość w formacie PDF ]