A couple of notes on the Ethernet Basics chapter from the CCIE Routing and Switching Exam Certification Guide

Wiring – On Routers & PCs transmit pair is wires 1 & 2 and the receive pair is 3 & 6, switches are the opposite way around this is so the receive pair will communicate with the transmit pair between devices. If two PCs (or routers or switches) are directly connected, to ensure the transmit and receive pairs match up a crossover cable should be used which swaps wire 1-3 and 2-6, alternatively the port will need to support Auto MDIX which can resolve the problem automatically.

Auto-negotiation – Fast Link Pulses are used to detect the speed of a connected interface with 17  100ns pulses with informs the  device on the other end of the wire the speed/duplex settings. Fiber does not use Fast Link Pulses but the negotiation works in a similar way.

Preamble - The preamble consists of 62 alternating 1s and 0s ending with a pair of 1s, this 8 bytes of signalling information is sometimes referenced as a 7byte preamble and 1 byte start of frame delimiter.

U/G, U/L MAC Bits – I wrote a post a while ago regarding a couple of bits in the MAC, its important to note that the second from last bit in the first byte of the MAC is the Individual/Group (U/G) bit which if set to 1 signifies that its a multicast or broadcast MAC. Also the last bit is Universal/Local (U/L) bit which indicates where it is set to a 1 indicates it has been administratively as opposed to vendor assigned, but many devices and drivers do not enforce the U/L bit.

The command no ip proxy-arp was one of those things which I saw in IOS configs and wasn’t to sure what proxy arp is used for or why it exists. Proxy ARP is where a router will respond on behalf of another device, it was used heavily in networks before the days of DHCP & default gateways where a host would ARP for an address that wasn’t on its subnet (modern networks just send the packets to the default gateway instead of arping for the address), the router on the local network would then act as a “proxy” and respond on behalf of the device outside of the subnet.

Proxy ARP isn’t used if hosts are set with default gateways or have routing intelligence, setting a default gateway instead of using proxy ARP is a much better option. Using Proxy ARP instead of a default gateway results in higher ARP traffic & the ARP tables of the hosts get very large as they maintain an IP/MAC binding for every single address the communicate with.

A pop quiz fact: I haven’t been able to verify this from another source but somebody at work told me that IBM helped created DHCP as they would assign workers/visitors with static IP addresses and as people moved around addresses would get lost, eventually they lost 750,000 addresses and needed to create a solution which stopped this from worsening. This is hear say and I cant vouch for its authenticity but its a decent story to talk about the need for DHCP.

Today I found out another interesting bit of information about the MAC address, actually its not just 1 bit its 2 bits.

The most significant byte is the first byte in the address and in that byte the last bit is the most significant bit.

Also the most significant bit is called the Individual/Group (I/G) bit and the next most significant bit is the Universal/Local (U/L) bit.

If the I/G bit is set to 0 it defines that the address is unicast and conversely if it is set to 1 it is multicast or broadcast.

If the U/L bit is set to 0 it means the address is vendor assigned but it if it set to 1 it is administratively assigned, its noted tha many devices do not enforce the rule of the U/L bit.

Automatic Medium-Dependant Interface Crossover is a licenced technology from HP. According to their website at http://www.hp.com/hpinfo/abouthp/iplicensing/automdix.html 200million ports supporting the technology will ship this year.

Its not supported on all Cisco switches, & it prevents the wrong cable from being installed, eg a striaight through between two switches, by simply swapping the transmit and receive pairs if the wrong cable has been used.

Chapter 1 on Ethernet mentioned about autonegotation using Fast Link Pulses so I wanted to find a little bit more about how this works. Cisco devices detect link speed and duplex settings using Fast Link Pulses (FLP) which is adapted from Normal Link Pulses (NLP) defined in the 10BASET specifications. This negotiation will only occur over 8 wire twisted pair copper and as such wont happen on other media types such as Fiber.

Upon an interface coming up a FLP burst will send 17 pulses of 100ns each representing a Link Code Word (LCW) which informs the other end of the link about the media speed and type and standards to use. The LCW only sends enough information to setup link speed of 100Mbps, so for Gigabit speeds multiple LCWs are sent which contain extra information such as master/slave settings, and multiport capabilities.

Although other media types such as fiber don’t use FLP they do use similar auto negotiation protocols.

Before FLP was defined devices would send an idle signal down the wire and the other end would train to it this is called parallel detection.

Its really important to note that the 802.3u standard does not specify what FLP should do if the link has statically defined speed/duplex settings, so it reverts to parallel detection and correctly sets the speed but assumes the link it half duplex. This results in a duplex mismatch if one end of link is coded for 100Mbps Full and the other for autonegotiate.