Routers -vs- Switches?

poochie asks "I'm currently in high school taking the Cisco first year course (semester 1 and 2). We're messing around with routers, and I'm starting to get the idea that routers are being outdated. We're learning switches next semester, but from what our instructor says, switches are just more specific to a LAN. In my mind, they seem more versatile in that they actually control their own ports, rather than having a router which is configured to support lans with switches on them, which have to be configured anyway, and make everything more confusing. I dunno about the purpose or the limitations of each, but from what I see, routers make life difficult for small networks. So I guess my question is twofold: Could anyone list resources on the Web about information on routers vs switches, and does anyone see switches taking over routers/hubs in the future, and making network administration a lot easier? "

Comments

ask "The Internet company"

[ptbrown]

Cisco has a smegload of documentation online. It's mostly product-specific, but some very good general information can also be found.

Here' s the answer to your question.

The fundamental difference between a LAN switch and a router is that the LAN switch operates at Layer 2 of the OSI model and the router operates at Layer 3. This difference affects the way that LAN switches and routers respond to network traffic.

and then, just beneath that

Because routers implement Layer 2 functionality and switches are beginning to implement Layer 3 functionality, the functions of a LAN swith and a router are merging.

This reminds me of the OSI Seven-Layer Burrito. Unfortunately, it appears to have vanished from the internet. c'est la vie.

Re:ask "The Internet company"

[eveso]

That would be a *switch* implements layer 2 functionality, and beginning to implement layer 3 functionality. A popular definition is that a Layer 3 switch is a router that routes more than 1.5 million PPS (packets per second). Another is that a Layer 3 switch is a router that routes in hardware instead of software. Now, the main issue here is that you really need both: a layer 2 device creates collision domains, while a layer 3 device lets traffic flow between subnets. Okay, this is a HUGE generalization, but hopefully you get the idea.

Re:ask "The Internet company"

[synaptic-impulse]

great explaination - but I dont think that this kid is quite to the level to understand what you just said.... he appears to not know the diff. between routers and switches.

The term "Switch" is abused

[Paul Carver]

A layer 2 (i.e. Ethernet) switch is stretching the definition and a layer 3 switch is just marketing hype.

A switch is a circuit oriented device. A switch such as the 4E, 5E, or DMS receives a connection initiation and responds by connecting an physical source port to a physical destination port. Everything coming into the source port goes right out the destination port.

The so called layer 2 and 3 switches are nothing of the kind. They receive individual packets or frames and examine them and decide which output port each should be sent to. I don't care whether they're examining the IP address, MAC address, or some other "tag". The function they're doing is routing.

I will grudgingly grant you the term Ethernet switch because each port on an Ethernet switch is connected to a device with a specific MAC address. You can sort of pretend that each Ethernet frame is a mini call and the switch is establishing a circuit from source to destination for the duration of a single frame.

A layer 3 switch is really a router. I don't care how many ASICs it has, it looks at each packet and sends it in the general direction of its destination based on information derived from routing protocols.

Re:The term "Switch" is abused

[Techno_Jesus]

I sooo agree... I've gotten into many a fight with cisco consultants about the theory behind switches... They are in fact Layer 2 (i.e. ethernet) routers and are even more so routers when you have a store and forward vs. a cut through switch. Sure it doesn't fit the tradiitional definition of router (Layer 3+, IP, IPX, etc) but functionally it does the same thing... Take our network here, there is a cisco 3548 acting as a core switch with 8 cisco 2924M-XL's cascaded off of it (2 segment etherchannel to each) when a workstation wants to talk to a server (located on the 3548 as well) the ethernet frame is "routed" from it's switch accross the etherchannel to the core switch. The clients switch "decided" that it needed to go accross the channel to reach such-and-such MAC address... That my friend is routing, and spanning-tree is a routing protocol :) Not in the traditional sense of course...

-Aaron

Re:The term "Switch" is abused

[lazlo]

More accurately, what is currently called a switch is a multiport bridge. Well, actually an Ethernet switch is a multiport bridge. An ATM switch or Frame Relay switch is more like a real switch.

But, to be more informative, and to relate to the original question, routers will not be replaced by switches (read: bridges). Bridges theoretically have to know the address of every device addressable on their network (read: broadcast domain) Routers deal with layer 3 protocols which have (or at least should have) hierachical addressing. They do not have to keep a forwarding table for every device, only one for every subnetwork, network, or supernet. The current internet has on order of 80K routes. That stretches the capabilities of some routers. I know of few devices that could keep a forwarding table of a billion addresses (one quarter of the IPv4 address space.... which is running out.) and do anything usefull with it. What will happen (and is happening) is that routing functions will begin to be integrated with switching functions.

The way Cisco's "Layer 3 Switches" currently do things is this: The switch is connected to a Route Processor, usually a 7000 series router or an RSP blade. If a packet comes in that is destined for the MAC of the router, it's layer 3 destination is examined and placed into a "flow." When that packet comes back out of the router, that "flow" is completed, and further packets from the same port with the same destination get sent out the port that the original packet went out of, rewritten slightly to reflect what would have happened if it had gone through the router. This increases speed in two ways. First, the packet didn't have to put up with the latency in the router. Two, it only had to traverse the backplane once. There are tradeoffs, of course, but it is a nifty technology. Note, however, that you still need a router in this scenario. And when you don't need a router, it's because your switch is acting like a really stupid (but really fast) router.

Re:The term "Switch" is abused

[fm6]

Paul makes an important point about semantic inflation. Give him a higher score!

Re:The term "Switch" is abused

[exult]

Please excuse the interruption, but I'm looking for a router/switch engineer interested in doing some freelance work. If interested, please respond to redfishtrout@yahoo.com Thanks.

Re:The term "Switch" is abused

[Detritus]

There have been switches that switched packets, not circuits, for over 25 years, before IP or the OSI networking model even existed. They were called switches, not routers.

Re:High school?

[sid crimson]

Cisco in high school. That woulda been good.

I was happy to get Pascal *and* AP Pascal. Networking OS? Not on the Apple IIe. :-)

-sid

Re:High school?

[Techno_Jesus]

My old high school just recently got a grant or something to have a cisco academy... It's in the middle of no where with and under 400 student count and their lab for this is 30 seats.. That means they expect 7.5% of the high school to participate in this class every semester???? They are on crack, when I was there wasn't 7.5% of the high school that *wanted* to take typing, let alone the one class on VB... I hope they get someone who knows what they are doing because to this day their network is wide open on the net without a decent firewall... I've printed warnings on their printers but I guess they just don't get it.

Re:High school?

[fm6]

Let's face it, us geeks are no longer the High Priests of a Low Cult. Stats: computer ownership in the U.S. is doubling about once a year. That's impressive in itself, but what really is important is that most new users are low-income, non-college-educated people who have traditionally resisted using computers. (I'm regurgitating an NPR report that went out a couple months ago). Those of us who have been trading on are lack of technophobia for that last 30 years are going to have to start earning our keep!

Networking info

[Dalantech]

Go to the Networking section of my web page.

Re:High school?

[FattMattP]

You guys had computers in high school? Sheesh. I'm feeling old.

Re:High school?

[poochie]

The truth is, its great that i get courses like this offered to me. The problem is, the courses are in such demand that more and more teachers are egtting sucked into it and not nessecarily trained to the baseline. This problem leads to misinformed questions such as my own. I dont know if cisco will make much money off of this, as I am one of 5 or so students not failing (69 and below is failing) the class. Regardless, its a great attempt, and makes my life a bit more simple in the future.

Re:High school?

[def]

Currently, I'm a rising junior at Carnegie Mellon university, but my high school was the Academy for the Advancement of Scienct and Technology.

They've had some pretty stunning technology classes for a while, including classes in web design (a 4 trimester sequence!), and of course CS which they teach in C++. More recently (since I graduated), I'm told they have added classes in Cisco, complete with a lab that has about a dozen routers.

Their home page is http://www.bergen.org/AAST

ATM will take over eventually -- very eventually

[skids]

If your interested in following progess in WAN tech, then
check out the Fore Systems site at www.marconi.com.
ATM takes the best features of circuit based
and TDM networks and combines them with cell relay.
It is definitely the more intelligent solution,
however it won't be ubiquitus until all the brute
force methods people are using to speed up WAN
(like IP/WDM) eventually start failing to deliver
guaranteed Quality of Service as consuption increases.
See if you can find the "ATM academy" section of Fore's pages (seems it has moved since I was last there.)