routers??

hey Guys,

i just wanted to start this thread as i thought it could lead to an interesting discussion, as i have found it hard to find the answers to my questions on the internet. as, recently, i have become very interested in the intricacies of computer networks.

in particular i have become interested in the operation of routers, as these are the more complex networking devices, as apposed to hubs, bridges and switches. i am fairly confident that i understand the operation of basic LAN routers that are responcible for a subnet of a LAN.

therefore, i particularly wanted to discuss the operation of home network routers, or for that matter any router that uses NAT or any variation thereof. so, i was wondering if anyone could describe how NAT and PAT/NAPT work?


regards,

- user0182

http://en.wikipedia.org/wiki/Network...ss_translation

hey SirDice,

obviously, i have read this article. what i really need is to set up a LAN, where i can play with these things.

but, i was wondering about things like; if i were to write a client server application where the client program sent data to the server on a certain client port then the server read that data and responded to a different client port, and the client machine was sitting behind a router using NAPT Overloading, am i right in assuming that this would not work and confuse NAPT?

thus, any server that tries to send a response to a different port than the port that the request was sent from would fail where NAPT is involved? as a router using NAPT, maps client "IP Address/Port combinations" to it's own router ports.


- user0182

btw - i also wanted to ask what networking hardware you guys consider to be the best, Linksys, Belkin, Netgear, D-Link etc.

Quote:

---

obviously, i have read this article

---

Did you?

I gave the article a good read-through and under this heading Basic NAT and PAT is a link to Network Address Port Translation, NAPT which links to another article that breaks NAPT down fairly well.

In this linked article, I found this under Translation of the Endpoint:
"Otherwise, if the destination port number of the incoming packet is not found in the translation table, the packet is dropped or rejected because the PAT device doesn't know where to send it"

Which answers this:

Quote:

---

thus, any server that tries to send a response to a different port than the port that the request was sent from would fail where NAPT is involved?

---

Am I missing something here?

Quote:

---

Originally Posted by user0182

obviously, i have read this article. what i really need is to set up a LAN, where i can play with these things.

---

No need, install VMWare server. Install one or two basic bsd/linux VMs in it and you've got enough to play with..

Quote:

---

but, i was wondering about things like; if i were to write a client server application where the client program sent data to the server on a certain client port then the server read that data and responded to a different client port, and the client machine was sitting behind a router using NAPT Overloading, am i right in assuming that this would not work and confuse NAPT? thus, any server that tries to send a response to a different port than the port that the request was sent from would fail where NAPT is involved? as a router using NAPT, maps client "IP Address/Port combinations" to it's own router ports.

---

Look into how an active FTP works if you're behind NAT (or better why it doesn't work).

Quote:

---

btw - i also wanted to ask what networking hardware you guys consider to be the best, Linksys, Belkin, Netgear, D-Link etc.

---

Cisco.. But then.. We have to manage about 10.000+ routers and switches. That's a whole different ballgame compared to a run-of-the-mill home setup ;)

by "obviously", i just mean't that doesn't everyone check Wikipedia first? i did read the Wikipedia page, but i didn't read every article that it linked to.

i just wanted to check with you guys that i had understood what i had been reading, as i'm sure it is possible to read something and think that one has understood it, but still be mistaken.

i'll look into the FTP thing, SirDice, but i did read something about it. as i understand, FTP won't work throught NAT because the packets include IP Address information in the actual packet data, above the Network layer?

anywho, i was wondering what peoples opinions of networking hardware is and which is considered to be the best? i've been reading various product descriptions, for switches and routers, and at the moment i'm thinking that Linksys might be a good way to go, to run a multi-OS network (Linux, BSD, OS X, XP, Vista etc), what do you guys think?


regards,

- user0182

Quote:

---

anywho, i was wondering what peoples opinions of networking hardware is and which is considered to be the best

---

Well, we use a Cisco firewall and once you learn to configure it, it's a very solid solution. And for workgroup switches(for extending your network), Linksys products have worked great for us. As for main switches, I like the Cisco Catalyst Series.

The only thing is, you're dealing with a multi-platform, multi-vendor network. I deal with all Windows based platforms so I don't know how well everyone would play together under your network.

Networking equipment has feck all to do with what OS your clients/servers are running.
Routers/switches work with TCP/IP and are totally oblivious as to what OS you happen to run on your workstation/server.

Quote:

---

Networking equipment has feck all to do with what OS your clients/servers are running.
Routers/switches work with TCP/IP and are totally oblivious as to what OS you happen to run on your workstation/server

---

I was speaking more of a management perspective. As in, how different OS's interact with the equipment itself via GUI, 9-pin, etc. But hey, thanks for the Networking for Dummies lesson! :rolleyes:

Quote:

---

Originally Posted by ShagDevil

As in, how different OS's interact with the equipment itself via GUI, 9-pin, etc.

---

Not really OS dependent now is it? Most, if not all, GUIs on networking equipment are web based.
It works exactly the same on Freebsd, Linux, windows, whatever.. 9-pin? Ah.. Serial.. Same deal..

I have never ever seen network equipment that's only manageable if you use a certain OS.

The great thing about TCP/IP is... is it's OS independent.. So why make equipment that only works (manageable) with a certain OS?

Quote:

---

It works exactly the same on Freebsd, Linux, windows, whatever

---

I bet people who work with multi-vendor networks know that. One small mental note, I only work with Windows platforms.

Quote:

---

I have never ever seen network equipment that's only manageable if you use a certain OS

---

That's not really the point here. There's some pretty nice third-party switch management software that I've seen that may not run the same (or run at all) on hosting servers with different OS's (like OpManager). This is something that may be important to consider in a mixed-vendor environment depending on what you're looking for.

Quote:

---

The great thing about TCP/IP is... is it's OS independent

---

:wow: No way!

Quote:

---

Originally Posted by ShagDevil

:wow: No way!

---

Way :D

There is a lot of good information on the Security Now podcast. Episode 3 and 42 explains a lot about NATs and how they work ( http://www.twit.tv/sn3 http://www.twit.tv/sn42 ). There are also many other basics in the early podcasts that can help you understand how different networking protocols work. Here is the episode guide complete with transcripts: http://www.grc.com/SecurityNow.htm . It's really easy to follow and I learned a lot myself from these podcasts.

also, another thing that i was wondering about was; can a network switch be connected to another network switch?

i would imagine that there are 2 possibilities. either a switch expects each of its connections to be a specific IP address, and so connecting 2 switches togther would not work. or the two switches would work together, acting as though their is a network bridge between them. i'm expecting that my second assumption is likely to be closer to the truth.


regards,

- user0182

Quote:

---

Originally Posted by user0182

also, another thing that i was wondering about was; can a network switch be connected to another network switch?

---

Yes. Most switches (even the cheap ones) have an "uplink" port. Use that port to "uplink" it to another switch.

So you get something like this:

Code:

---

[router]-----{uplink}[switch1]{P1}-------{uplink}[switch2]---<PCs>
                            \{P2}-------{uplink}[switch3]---<PCs>

---

why is there a distinction between the "uplink" port and the other ports? i guess that there may be a reason to use an "uplink" port to connect to a router, although i still don't fully understand.

if i understand correctly; a switch is just a series of network bridges, and in principle is the same as placing a bridge between each of the machines that you have attached to the switch.

therefore, as a bridge learns which IP addresses that are on either side of it's interfaces, shouldn't any switch be able to learn a group of IP addresses for each of it's interfaces?

thus, wouldn't the following work?:

Code:

---

[router]----(uplink)[switch](p1)----(p1)[switch2](p2-p4)----<PCs>
                          /(p2)----(p1)[switch3](p2-p4)----<PCs>

---

although, i still don't really understand why the following would not work?:

Code:

---

[router]----(p1)[switch1](p2)----(p1)[switch]----<PCs>
                        /(p3)----(p1)[switch]----<PCs>

---

regards,

- user0182

Quote:

---

Originally Posted by user0182

why is there a distinction between the "uplink" port and the other ports?

---

The RxD and TxD are reversed.
Having said that, a lot of modern switches/networkcards are auto-sensing and switch automatically from "normal" to "uplink" (i.e. reversing TxD/RxD).
With the more expensive equipment the "uplink" port is usually also faster. So for instance you'll have 24 10/100Mbit switched ports and 2 1Gbit uplinks.

Without this automatic changing, if you need to connect:

[router]-----{uplink}[switch] (straight cable)
[switch]{p1}----[PC] (straight cable)
[switch]{p1}----{uplink}[switch2] (straight cable)
[switch]{p1}--X--{p1}[switch2] (cross-cable)
[PC]--X--[PC] (cross-cable)
[router]--X--[router] (cross-cable)

Quote:

---

therefore, as a bridge learns which IP addresses {...}

---

A bridge works on layer 2, IP is layer 3.

what is RxD and TxD?

what i meant by a bridge learns IP addresses, is it learns MAC addresses but as each MAC address maps to an IP address, therefore it sort of learns IP addresses.

Quote:

---

Originally Posted by user0182

what is RxD and TxD?

---

Receive Data and Transmit Data.. Of the 8 pins in a RJ45, 4 are connected (with 'regular' 10/100Mbit ethernet).
2 for RxD and 2 for TxD. You connect a Transmit to a Receive and vise versa.

Quote:

---

what i meant by a bridge learns IP addresses, is it learns MAC addresses but as each MAC address maps to an IP address, therefore it sort of learns IP addresses.

---

A bridge has no notion of the MAC-IP relationship. Do learn the correct basics or you'll end up not understanding the more complex stuff ;)

http://en.wikipedia.org/wiki/OSI_model

Quote:

---

A bridge has no notion of the MAC-IP relationship. Do learn the correct basics or you'll end up not understanding the more complex stuff

---

now i am confused! i know that a standard switch does not look at, nor has any understanding of IP addresses, i just think i may not have been communicating accurately, as most of the time sending to an IP address or a MAC address on a LAN has the same effect. i thought that a switch is a series of bridges, and each bridge learns the MAC addresses that are on either side of itself (i.e. operating at layer 2 of OSI), and thus only transmits packets where the source and destination MAC addresses are on different sides of the bridge. am i correct?

btw: i really appreciate all of your help SirDice. this is why i wanted to create this thread, as i find this stuff interesting and i wanted to check my knowledge.


regards,

- user0182

- user0182

Quote:

---

Originally Posted by user0182

as most of the time sending to an IP address or a MAC address on a LAN has the same effect.

---

Try adding a router to the equation. Look what happens to the MAC addresses once a packet crosses a router.

Quote:

---

i thought that a switch is a series of bridges, and each bridge learns the MAC addresses that are on either side of itself (i.e. operating at layer 2 of OSI), and thus only transmits packets where the source and destination MAC addresses are on different sides of the bridge. am i correct?

---

That sounds about right.

i think that basically my question is; can a switch deal with more than one MAC address connected to it's interfaces?

for instance; if one were to connect a hub to a switch or another switch to a switch.

so, i am also wondering what is the difference between a switch's normal interfaces and an "uplink" interface?

Quote:

---

Originally Posted by user0182

i think that basically my question is; can a switch deal with more than one MAC address connected to it's interfaces?

---

Yes, it can. They're stored in what's known as a CAM table.
http://en.wikipedia.org/wiki/CAM_Table

Quote:

---

so, i am also wondering what is the difference between a switch's normal interfaces and an "uplink" interface?

---

As I've said before, RxD and TxD are reversed and for the more expensive switches the speed is usually higher (1Gbit uplink vs. 100Mbit ports).

i still don't quite understand. i am assuming that by "RxD and TxD are reversed"; you mean that the actual wires in the socket are reversed?

if so, i don't understand why the wires have to be reversed, could you explain?


regards,

- user0182

Normally TxD is pin 1 and 2 and RxD is on 3 and 6.

1) TxD+
2) TxD-
3) RxD+
4)
5)
6) RxD-
7)
8)

What would happen if you used a straight (1-to-1) cable to connect two similar ports together?
RxD on one side will be connected to RxD on the other. Same for TxD. This won't work.

If RxD and TxD are reversed RxD on the one side will be connected to TxD on the other and vise versa.
So either RxD/TxD need to be reversed on the device itself or you'll need to use a crosscable.

okay, so i now understand where crossover cables come into the equation.

so now let me just test my knowledge. if, for instance, you have a standard switch with only 4 standard interfaces and no uplink interface, and you connect 4 clients to a hub and connect the hub to one of the switch's interfaces (all with straight cables), then you connect 2 clients to 2 of the switch's other interfaces (with straight cables), and finally connect the remaining switch interface to another switch's standard interface with a crossover cable. does this sound as though it would probably work?

btw: is there an easy way to tell different port types apart (i.e. standard from uplink)?


regards,

- user0182

If the hub doesn't have an uplink port you'll need to use a crosscable to connect it to the switch.

And never forget to set the switch port to half duplex if you connect a hub to it!

Uplink port is usually marked as such or with an X.. You find out quick enough as the link led won't come on..

thanks SirDice, that has been really helpful, i think that my knowledge is getting much better. i was also wondering, is there a convention from marking auto-sensing ports, or is it just a matter of reading the documentation?

i was also wondering; when one has a router that is connected to the Internet, how does that router know where to route packets that are not addressed to IP addresses on one's LAN? does one's router send all of these packets to a default IP address of a router at one's ISP?


regards,

- user0182

Quote:

---

Originally Posted by user0182

i was also wondering, is there a convention from marking auto-sensing ports, or is it just a matter of reading the documentation?

---

RTFM works for me ;)

Quote:

---

i was also wondering; when one has a router that is connected to the Internet, how does that router know where to route packets that are not addressed to IP addresses on one's LAN?

---

The (static) routing table defines were to route things. Try (on windows or *nix) netstat -rn.

There can be some dynamic routing too. Things like RIP, OSPF and BGP.

sorry, i still don't understand. so what process does the router go through for packets which are addressed to IP addresses that are not on the router's LAN (e.g. accessable via the Internet), and the router it's own connection to the ISP?

You'll need to understand about ip addresses and subnet masks. Especially what's the network and host address of an IP.

The router will look at the destination IP and compare that to it's routing table. If it cannot find a match it will send it to the default gateway.

The next router (usually on your ISPs network) will do the same, and the next router, and the next until the end point is reached.

yeah, i do have a pretty good understanding of routing tables.

i think that answered my question. i know that, as far as machines on one's LAN are concerned, one's router (which has an Internet connection) is the default gateway. i was just wondering how that router routes packets?

so it too has a default gateway, which is a router at one's ISP, where packets with a destination that is not on one's LAN are routed. so, for example, if you were to request a web-page, the HTTP request would probably be sent to the router on your LAN, then your ISPs router, then an Internet backbone router etc. am i correct?


regards,

- user0182

Quote:

---

Originally Posted by user0182

i think that answered my question. i know that, as far as machines on one's LAN are concerned, one's router (which has an Internet connection) is the default gateway. i was just wondering how that router routes packets?

so it too has a default gateway, which is a router at one's ISP, where packets with a destination that is not on one's LAN are routed. so, for example, if you were to request a web-page, the HTTP request would probably be sent to the router on your LAN, then your ISPs router, then an Internet backbone router etc. am i correct?

---

Yep, that's it. Each router forwards it to the next depending on it's routing table.

On your PC and/or home network router the routing table is usually pretty simple, 1 or 2 local networks and a default gateway. The routing tables on the backbone of a big ISP can get pretty large. ISPs usually don't have static routes but use BGP to dynamically build the routing tables. With BGP routers can talk to each other so information about what network is reachable via what route gets automatically distributed.

so, how does a home router figure-out what it's default gateway should be (i.e. how does one's LAN router get the IP address of the ISP's router)?

A home router usually either a) gets an IP from the ISP automatically via DHCP, which is leased to the router's WAN port, or b) has the WAN IP set manually, either by instructions sent from the ISP or other means. Usually, when DHCP is involved, not only is the IP for the router itself sent, but also other information, such as the IP of the default gateway, IP(s) of the DNS server(s), IP(s) of the DHCP server(s), lease time, etc...

NukEvil is correct.

You either get the default gateway from DHCP or you set it by hand.

thanks guys, that's been really helpful and interesting.

so, does that mean that a typical home router acts differently on it's different interface types (i.e. uplink, standard). i mean; as many home routers act as a DHCP router for their LAN nodes, does this mean that the router acts as a router/server on it's standard interfaces and as a client on it's uplink interface?


many thanks,

- user0182

If it's set up that way, yes. Most home routers are a DHCP client on the WAN interface and a DHCP server for their internal (W)LAN interface(s).