I read this chapter through in a sitting and didn’t spot much I wasn’t already comfortable with. Although there’s a lot of details that you just don’t think about much that come back up like heartburn. Chapter highlights:
- Ethernet pinouts. 568A vs. 568B, and the differences between a straight-through and a crossover. I was disappointing that they are stuck in CAT5-land. We just built out a new data center with CAT6E and CAT6 augmented. CAT6 compliant xover cables cross pairs 4-5 with 7-8 as well as the 1-2 with 3-6 that we know and love. It also became obvious in Figure 1-1 that their diagrams were prepared in color, and then translated to grayscale with no legend. While I figure we all know 568A & B color-schemes and pinouts with no issue, a lack of color for a picture like that makes the OECG an incomplete reference.
- Auto-MDIX mentioned. Can’t say I’m a fan of auto-mdix.
- Good review of speed, duplex and auto-negotiation issues. I picked up a good detail here – when the switch is in auto-neg and the other side isn’t, the switch will default to half-duplex for fast and full-duplex for gig. The speed can still auto-net successfully based on the sensing of the electrical signalling coming into the switch port. But duplex can’t be detected unless both side are doing auto-neg. I didn’t know that the reason the switch settled on 100/half or 1000/full respectively was because of a default setting.
- CSMA/CD. Carrier sense multiple access with collision detection. That part of ethernet hasn’t changed much. And let’s face it – it’s the inefficiencies of a shared medium like this that hastened switching technology. The switch saved ethernet. Broadcast domains and collision domains were also discussed.
- Another interesting point I didn’t know. If you’re stuck using a hub, the hub’s uplink port on a switch should always be half-duplex. Since a hub is a simple electrical repeater, you’ll get collisions from time to time, so you need to have a half-duplex uplink from the switch to the hub. You should only do full-duplex on segments where there can’t possibly be a collision. I need to think more about the “why” of this.
- Commands for speed & duplex are discussed.
- There’s an IOS section that mentioned, among many other things, the CDP error that will log when the CDP process detects that there’s a duplex-mismatch between himself and his neighbor.
- Moving on to Layer 2 now, there was a section on ethernet framing. DIX versus Original IEEE 802.3 versus IEEE 802.3 with SNAP. And it’s labeled as a “Key Point”, meaning that even if you never use this in real life, you’ll need to know it for the test. I can’t think of too many situations where I’ve needed to know ethernet frame header fields, but I guess it can’t hurt. Okay, here we go.
Preamble (DIX) – it’s for synchronization and clocking. Alternating 1’s and 0’s, ending in a pair of 1’s. 8 bytes.
Preamble and Start of Frame Delimiter (802.3) – the same purpose as above, only this gets used in the non-DIX frame types and it’s broken out as 7 bytes and 1 byte.
Type (DIX) – tells you what type of L3 protocol follows the header. 2 bytes.
Length (802.3) – byte length of data following the length field, up to the trailer. So the receiver knows when the end is coming…which, let’s face it….we’d all like to know when the end is coming, right? 2 bytes.
DSAP Destination Service Access Point (802.2) – protocol type field, 1 byte.
SSAP Source Service Access Point (802.2) – describes upper layer protocol that created the frame, 1 byte.
Control (802.2) – provides mechanisms for connectionless and connection-oriented operation. Not used much, 1 or 2 bytes.
Organizationally Unique Identifier (SNAP) – generally unused today, 3 bytes.
Type (SNAP) – Uses same value as the DIX type field, and overcomes the shortcomings of the DSAP field, 2 bytes. - In the context of ethernet (not IP), unicast, multicast and broadcast addressing are discussed.
- Ethernet addresses are 6 bytes: 3 uniquely assigned to a vendor by IEEE and 3 uniquely assigned to by the vendor to the device. In theory, every ethernet device in world has a unique MAC.
- Hopeless detail that just might crop up. Ethernet addresses have their most significant BYTE on the left. But inside each byte, the BITS are opposite – the leftmost bit is the least significant, known as noncanonical or little-endian. Why do we care? Because the 2 most significant bits of the first byte (i.e. the right hand ones) are special. The U/L bit (position 7 ) and the I/G bit (position 8) matter to us.
If the U/L (universal/local) bit is 0, it means the address is vendor assigned; if U/L is 1, the address has been administrative assigned and overrides the vendor value.
If the I/G (individual/group) bit is 0, the address is a unicast. If it’s 1, the address is broadcast or multicast. - A nice section on switching and bridging logic, reviewing what a switch does as certain frames come inbound.
Known Unicast – switched to the port where it’s known
Unknown Unicast – floods to all ports, except where it showed up to begin with
Broadcast – floods to all ports
Multicast – floods to all ports, unless you’re managing multicasts with IGMP or something similar. (I don’t know what other than IGMP might do this, it just popped in my head. The book doesn’t say.) - IOS code section reviewing the show mac-address-table commands.
- The end of each chapter has a “Foundation Summary” section that is NOT a review. Rather, it’s a dumping ground of all the stuff they felt were worth mentioning, but didn’t need an individual section devoted to them. Hooray. So…here’s more info from the Foundation Summary.
- Ethernet alphabet soup: 10Base5/thicknet, 10Base2/thinnet, 10BaseT, DIX Ethernet v2 DIX=Digital/Intel/Xerox original spec for ethernet, IEEE 802.3 media access control specification, IEEE 802.2 logical link control specification, IEEE 802.3u FastE spec, IEEE 802.3z GigE over fiber spec (ZIPPY), 802.3ab GigE over copper spec.
- Switching methods: store-and-forward (get it all, check FCS, then forward), cut-through (get the destination address, then forward before the rest of the frame shows up, low latency), fragment-free (get the first 64 bytes, then forward before the rest of the frame shows up, low latency, plus assumes a collision didn’t happen since that should have been detected before 64 bytes showed up).
- Various IOS commands: interface, duplex, speed, show mac-address-table, show interface x, show interface vlan x.
- UTP cabling- Cat1 (telephone only), Cat2 (4Mbps for token ring over UTP), Cat3 (10Mbps), Cat4 (16Mbps for fast token ring), Cat5 (1Gbps), Cat5e (1Gbps, really to support GigE over copper), Cat6 (1Gbps+, support for multi-gigabit speeds)
- Ethernet cabling types and max run length:
10Base5 – thick coax – 500m
10Base2 – thin coax – 185m
10Base-T – UTP Cat3/4/5/5e/6 – 100m
100Base-FX – MMF – 400m
100Base-T – UTP Cat3/4/5/5e/6 – 100m
100Base-T4 – UTP Cat3/4/5/5e/6 using 4 pair instead of 2 – 100m
100Base-TX – UTP Cat3/4/5/5e/6 – 100m
1000Base-LX – MMF or SMF – 3Km MMF/10Km SMF
1000Base-SX – MMF – 220m 62.5micron/550m 50micron
1000Base-ZX – SMF – 100Km
1000Base-CS – STP 2 pair – 25m (gee, we see LOTS of that in the real world)
1000Base-T – UTP Cat5/5e/6 4 pairs – 100m