Local Area Networks

1. Local Area Networks LAN

2. Why LANs? • Provide a means of DIRECT connection to other machines • Manage access • Provide reasonable performance • Hopefully allow for interconnection with other LANs

3. LAN criteria • Resolve access • Fair access • Quick access • Fast transmission • Security • Robustness

4. Do LANs meet these criteria? • There are a huge number of network designs which have been proposed • Few are actually in operation • Most LAN designs excel in a few of the previous areas but not all • OSI allows for integration of newer technologies as they evolve. • For now-> price & performance drive us to ethernet.

5. Ethernet What you are likely to see

6. Ethernet • Broadcast medium • Everyone sees all transmissions* • Orderly free-for-all to determine access • Cheap • High performance (few Mbps to Gbps) • Many standards • Speed • Medium

7. Basic Principles • Access strategies vary significantly. • Centralized • Some main controlling unit deciding • Less Robust • Decentralized • Group decisions • Controlled behavior • Moving centralized control

8. Other classifications • Reserved vs non-reserved • Static versus dynamic • Broadband (FDM) vs baseband (TDM-ethernet) • LAN, MAN, WAN • Optical vs electrical

9. What strategy does ethernet use? • Less time required to decide access means faster access and less delay. • Simple algorithm like using a “party-line” in a telephone system • Pick up the phone and see if it’s in use • If not send • checking to make sure no one else did the same

10. Assume each at extreme ends A begins transmission (a<<1) A B A message almost reaches B and B starts A B B sees the collision A B A sees the collision A B After “2a”, either collision or not

11. And what does that tell us? • Within “2a” frame transmission times, the initiator will • See a collision and stop transmitting • See NO collision and proceed with the assurance that NO ONE ELSE will transmit as others will look before beginning. • But what happens after a collision? • No central organizer • Wait a random amount of time and try again • If another collision -> WAIT LONGER!

12. How many collisions can occur? • Eventually the physical layer responds that it can’t deliver and the (OSI) layer above decides what to do next • There is NO UPPER BOUND on how long it will take to get access • Can’t service some types of traffic well • Video and voice

13. How does it work so well? • If “a” is small enough, a number of contention cycles (“2a”) can occur for each transmission. • E.g. if 2a= .002 and it takes 5 collision cycles, the overhead is 5x.002=.01 of the time and 1/(1+.01) is still very good. • Capacity typically far exceeds demand which means most of the time the network is free and no collisions occur. • Degrades in a heavily loaded environment!

14. What does it look like? • Originally a bus architecture • Now a logical bus but a physical star HUB computer computer computer computer

15. SWITCH computer computer computer computer SWITCH computer computer computer computer What impact does that have on performance? What impact does that have on network security?

16. Queueing Revisited Offered Load Versus Throughput Throughput ideal more typical 100% ethernet 100% 200% Offered Load

17. Token Rings An alternative

18. token In order to use the net, you must seize the token

19. message token

20. message token

21. message message token

22. token

23. What is the improvement? • A little delay in getting access • An upper bound on access as long as all behave well (and they do .. Same program) • Used in environments where upper bound is required like in automated manufacturing. • Previous queueing graph is more like the ideal, does not degrade to zero.