COMPUTER AND NETWORK
SYSTEM ADMINISTRATION
Summer 1996 - Lesson 14
Network Hardware
A. Ethernet - the dominant network solution
1. Broadcast protocol over a common wire
2. CSMA/CD - carrier-sense, multiple-access, collison-detection
3. everybody sends, if there is a collision, then back-off and try
again
4. for (1 <= N <= 10) pick a random number B between 0 and 2^(N-1)
wait for B * 51.2 microseconds and rebroadcast
for (11 <= N <= 15) max out at 2^10
for N = 16, give up
5. above a certain load the algorithm breaks down (some say
more than 30% saturation)
B. Ethernet cable types
1. 10base5 "thicknet"
- 50-ohm RG-11 coaxial cable
- N-type connectors
- vampire tap tranceiver or in-line
- can span up to 500 meters
- unwieldy
- making a good tap is a fine art
- might not be near a 2.5 meter "black mark" (especially in a lab
with many machines)
- connect to machine with an AUI (Attachment Unit Interface) cable
2. 10base2 "thinnet"
- 50-ohm RG-58 cable (close to cable TV type cable)
- BNC-type connectors
- use a T-connector to attach transceiver
- many computer and xterms have built-in tranceivers
- cable is more flexible
- easier to attach
- but maximum length is only 1/3 of thick coax
- if the machine is connected directly then must loop
into office (using up more precious length)
3. 10baseT "twisted pair"
- twisted pair cable with RJ-45 connectors
- star configuration to a "hub"
- easier to isolate problems since a machine will
only bring down its own wire
- versus a coax bus configuration where a bad transceiver or
connector can bring down everybody on the wire
- a bad hub can be a real pain, though, and is a single
point of failure
- start configuration is easy to install in a lab
but more difficult down a long hallways of offices
- length limit is 100 meters
- if having it installed in the walls then go ahead and
use Level V (category V) wire
4a. 100baseT
- "fast ethernet" (100 Mbits/sec) over same "cat 5"/RJ45 wiring scheme
as 10baseT
5. 10baseF "fiber"
- connectors, transceivers, and hiring someone to polish
and test the fiber is more expensive
- can extend your LAN up to 2 kilometers
- also, resistant to lightning
Here is an excellent source of Ethernet info.
C. Connecting and expanding networks
1. Repeaters
- operate at the physical layer of the network model
- do not recognize MAC addresses, IP numbers, or machine names
- they simply re-energize the packet and send it along
- multi-port repeaters offer more functionality
- they can isolate segments in addition to extending length
- Topology restraints (Exhibit E on page 299): two farthest points
must never be more than four repeaters apart
2. Bridges
- useful for reducing traffic load
- operate at the data link layer
- read the ethernet header
- know about what? (MAC addresses and packet types)
- bridges can be managed or unmanaged
- bridges build a table of MAC addresses to either pass through
or discard ("smart bridges")
- Question: say you have a host from which you are being
attacked (for example, daemon.mit.edu). Can you use the
bridge to lock out the machine from your net?
(no, only the MAC address of the next link in the route)
3. Routers
- operate at the network layer
- are protocol specific
- an IP router maps IP numbers to networks
- this is fine if you are only using IP on your net
- routers keep out unwanted traffic based on IP address
- the FSU routers use RIP; someday will use higher-level
interior and exterior gateway protocols
D. Other type of networks
FDDI
----
1. FDDI - Fiber Distributed Data Interface
- 100 Mb/sec token ring (scales better than ethernet under heavy loads)
- Can have a single ring or a dual ring
- Traditionally carried over fiber, also runs over
"cat 5" UTP (unshielded twisted pair) using RJ45 connectors
- Example: SCRI has an FDDI backbone for file servers and
ethernet bridges
- Emergence of "fast ethernet" eroding FDDI market?
(even though FDDI has better throughput with heavy loads)
FCS
---
1. FCS - Fiber Channel Standard
- 256 Mb/sec to 1 Gb/sec
- Point to point or a hub strategy
- Tiny part of networking marketplace; will probably only
be a high speed bus extension methodology
ATM
---
1. ATM - Asynchronous Transfer Mode
- book is somewhat sarcastic about ATM Here's a nice starting point to learn more about ATM.
- formidable objective:
A universal switching and multiplexing technique to
support integrated transport of multi-rate traffic
- capability for real-time transmission of voice, video, plus
high-speed data
- Data rates: OC-1 (51 megabits-per-seond) to OC-48 (2.488 gigabits-per-second!)
OC-3 (155 MBsec) is common.
- the entire issue of the Feb (1995?) Communications of the ACM is devoted
to ATM
2. other requirements
- must be cost-effective and scalable
> can't require a $5,000 device at every user interface
> but must scale up to devices that can handle large sites
- scalability is enhanced by a switch-based architecture
and a common cell structure
- must support multi-casting (one sender, multiple receivers)
3. driving applications
- digital medical imaging
- entertainment ("video on demand")
- supercomputer data transfers
- distributed network computing
4. ATM cells
- based on small fixed size cell
- 5 byte header
- followed by data segment
> much squabbling over proper cell size
> phone company wanted small cell to reduce delay (for voice)
> data folks want larger cell to reduce the amount of segmentation
and reassembly
- compromised on 48-byte data segment
5. ATM is connection-oriented
- before data is transferred a connection is requested
- as the connection is established VCI/VPI pairs are
allocated along the connection path
- the switches along the way allocate bandwidth and maintain
VCI/VPI mappings
- since a single path is allocated, cell ordering can be guaranteed
- because of low bit-error rates in optical fiber the error checking
only has to be performed at network-boundary nodes or at end-user
sites
E. Computer science LAN configuration
1. Balancing several factors
- maximum cable length (167 meters for thinnet)
- maximum number of machines (30 on a single segment)
- network load
2. Deal with length problems by using multi-port repeaters
3. Deal with number of machines per segment by using
multi-port repeaters
4. Deal with local congestion by using bridges
5. Router exists for outside world connection
F. software tools
1. ping
- in our 4-layer model, which layers must be functioning in in
order for ping to work?
- can you ping a machine which is in single-user mode?
- can you ping the PCs that run PCNFS?
- can you telnet to a PC running PCNFS
2. netstat -i
shows the machines network interface configuration
Name Mtu Net/Dest Address Ipkts Ierrs Opkts Oerrs Collis
le0 1500 128.186.0.0 mu 5726725 0 2037668 0 5830
lo0 1536 loopback localhost 183803 0 183803 0 0
- MTU: maximum transmission unit
> ethernet MTU is 1500 bytes
> this is the size of the packet not including the 14-byte ethernet
header or the 4-byte etherneet trailer
> a typical token-ring MTU is 4464 bytes
> a typical FDDI MTU is 4352 bytes
> a typical FCS MTU is 65280 bytes
3. netstat -r
netstat -r
Routing tables
Destination Gateway Flags Refcnt Use Interface
128.186.143.0 mpr.flhigh.fsu.edu UGH 0 0 le0
128.186.27.0 mpr.flhigh.fsu.edu UGH 0 0 le0
128.186.152.0 mpr.nursing.fsu.edu UGH 0 0 le0
localhost localhost UH 2 463210 lo0
192.239.192.0 scl-hub-fddi.fsu.edu UG 0 0 le0
default scl-hub-fddi.fsu.edu UG 1 397884 le0
192.239.193.0 scl-hub-fddi.fsu.edu UG 0 0 le0
146.201.0.0 scl-hub-fddi.fsu.edu UG 0 20848 le0
128.186.0.0 nu U 46 9180392 le0
flags - U: route is up
G: route is to a gateway
refcnt: current number of active uses per route
use: number of packets sent per route
4. netstat -a
netstat -a | grep login
rlogin to machine
netstat -a | grep login
5. traceroute
omicron:/usr/bin> traceroute gatech.edu
traceroute to gatech.edu (128.61.1.1), 30 hops max, 40 byte packets
1 scl-hub-fddi.fsu.edu (128.186.8.1) 3.741 ms
2 fddi-bfs.fsu.edu (128.186.254.5) 4.013 ms
3 SURAcisco-Firewall.fsu.edu (192.80.53.23) 4.336 ms
4 suranet.tlh.fl.us (198.102.72.10) 9.043 ms
5 atu2-tau1-c1.sura.net (128.167.125.1) 68.638 ms
6 git-atu2-c3.sura.net (128.167.160.2) 25.409 ms
7 gt-border.gatech.edu (192.221.26.1) 38.683 ms
8 gt-firewall-ext-fddi.gatech.edu (130.207.244.1) 24.813 ms
9 campus2-rtr-int-fddi.gatech.edu (130.207.254.3) 36.845 ms
10 gatech.edu (128.61.1.1) 40.945 ms
traceroute to berkeley.edu (128.32.123.6), 30 hops max, 40 byte packets
1 scl-hub-fddi.fsu.edu (128.186.8.1) 2.556 ms
2 fddi-bfs.fsu.edu (128.186.254.5) 3.641 ms
3 SURAcisco-Firewall.fsu.edu (192.80.53.23) 6.107 ms
4 suranet.tlh.fl.us (198.102.72.10) 7.811 ms
5 atu2-tau1-c1.sura.net (128.167.125.1) 58.469 ms
6 cpe1-fddi1.Atlanta.mci.net (192.221.42.100) 32.362 ms
7 border1-hssi1/0.Atlanta.mci.net (204.70.16.5) 35.987 ms
8 core-fddi-0.Atlanta.mci.net (204.70.2.49) 33.293 ms
9 core-hssi-4.Houston.mci.net (204.70.1.26) 42.221 ms
10 core-hssi-3.LosAngeles.mci.net (204.70.1.34) 87.002 ms
11 core-hssi-4.SanFrancisco.mci.net (204.70.1.42) 124.787 ms
12 border1-fddi0/0.SanFrancisco.mci.net (204.70.2.162) 137.09 ms
13 cpe1-hssi-1.SanFrancisco.mci.net (204.70.32.6) 145.582 ms
14 SU-CM.BARRNET.NET (192.31.48.200) 136.698 ms
15 UCB0.BARRNET.NET (131.119.2.2) 104.799 ms
16 inr-666-dmz.Berkeley.EDU (192.31.161.21) 231.589 ms
17 inr-108-styx.Berkeley.EDU (128.32.1.2) 127.806 ms
18 * inr-105.Berkeley.EDU (128.32.155.105) 87.194 ms
19 sunny.Berkeley.EDU (128.32.123.6) 151.383 ms