Science Fair Help

[ecumbee]

I am working on a science fair project on routing and switching. I need some help with ideas and resources, i recently completed the cisco networking academy program, after talking with my instructor our lab dose not have the capacity to handle a research project and regular classes. If anyone has any ideas please let me know.

[ahbanks]

It has been a decade since I've done a science project, but I'll give this a shot. Some clarification on what you're really looking for would be good, however.

General Ideas would include "How does VoIP work", "How does the internet work", "How do computers talk", "How do computers catch virii", "How does a hacker break into a computer".

The first three could be done in a research-type project. The last two could involve some programming, and would tie up a couple of PCs..

HTH....

[scottmac]

I'll toss in a couple ideas...depending on what level of complexity you're looking for.

The first isn't specifically switching / routing, but applies to virtually every network: The cabling system. You could set up varous cabling levels (cat3-6), you can show proper pair order (568a, 568b) versus random or "pair-pair-pair-pair" termination (results ina slit-pair on 3&6) ... there are a number of fairly easy demonstrations related to the most overlooked, most critical component of the network. There are also some fiber-related demos you could add.

Second, a bit more technical, would be to show the efficiency of CIDR blocking / route summerization over a typical scattered address networking. With a couple routers cranking out the routes, and a couple PCs running QCheck (www.netiq.com) you can show the end-to-end effect of good network design (better efficiency, better throughput, lower CPU utilization, etc).

Third, some demonstration of multicast (versus unicast) technologies (i.e., video over IP) and some of the associated switch/router multicast management systems (PIM, CGMP, IGMP, etc). Maybe tie that to some future thing like "Your next new TV station might be the Internet." Everyone can relate to TV, video makes a good "bug light" at shows, there's no real magic, but most people don't understand the systems and it looks like magic; it kind of depends on the crowd you're trying to address.

Good Luck

Scott

[ecumbee]

Those are some good ideas but there are several things i have to have

1. a Problem-something unproven or untried

2.ideas how to solve #1(hypothis)

3.Expermentation to prove #2

4.Data from experement to come up with #5

5.a conclusion did it or did it not work

I am still working on my ccna so i have very little experince in multicasting.

thanks for your help

[ahbanks]

Well, meeting criteria #1 may be difficult. The problems I'm thinking of, like how other traffic impacts VoIP or Video have some solutions.

How about writing your own routing protocol? Yes, several good algorithms exist, but they all have their problems.

How about developing methods for extreme-speed transmission over copper? I'm thinking like OC-192+ here...

Encryption methods are an always-evolving issue, as are distributed computing models

[ecumbee]

writing my own routing protocol?? that would be awsome it would realy show my high school instructor that i am not an idiot and that i can do things without cheating. could you explain more on the routing protocol

[ahbanks]

A routing protocol has two parts, IMHO. There is an algorithm to select the 'best' route to a destination, and a packet format to transmit that information to other devices (usually routers) that need it. In your case, I would pick a good protocol (EIGRP, IGRP, BGP, IS-IS, OSPF, with preferences to OSPF, IS-IS, and BGP), research how the protocol determines the 'best' route, and attempt to devise an algorithm that better models reality, is faster to compute, and be conservative on bandwidth. A good reference for protocols is Jeff Doyle's _Routing TCP/IP, Volume I_.

I'll use RIP as an example. When a router learns of a new route or destination via RIP, it will calculate what is the 'best' way to get there. In RIP's eyes, 'best' is the route with the fewest hops to get there. So, if there are two ways to go from A to B, one being a 56k circuit, the other an OC-48 connected to an OC-192, then RIP will pick the 56k circuit, since that is only one hop.

Obviously, improvements are needed.

For EIGRP, take a look at the Dijkstra algorithm. EIGRP uses bandwidth, delay, load, reliability, and MTU. I'm shooting off the cuff here, and you would need to enlist the help of someone who can do a reasonable job of multivariate statistics, but I would take a look at those metrics and see which ones are statistically valid. Then, I would start looking at other metrics to see if they would make any more sense. Essentially, asking the question, does the error rate on the interface impact performance more than delay?

I'll give this some more thought today and post some more later...