• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

If a regular router has switched ports,then why do they make "switch routers" ?

Quad

Golden Member
i'm a little confused.

please correct me if i'm wrong, but i'm under the impression that a regular dsl/cable router has switched ports. if this is the case, why do they make special "switch routers" that are more expensive?

is it because the dsl/cable router is not 10/100mb, and the "switch router" is?

thx in advance for clarifying
 
The "Switch" part of the consumer "Switch-Router" applies only to the Ethernet ports (all layer two stuff). When the packet is passed to the router part of the device, everything is happening like a traditional router: the packet is examined to see where it's going, where it came from, special handling (changing the address for NAT, etc). That process of examination and forwarding is done in software, and takes a fair amount of time.

The (much) more expensinve "Layer Three Switches" are also routers (by definition), but have specialized hardware (ASICs) for much faster examination, evaluation, and forwarding decisions. L3 switches are usually all single topology (like Ethernet), usually don't do serial (T1/PRI/56K/Frame Relay), and are designed for LAN segmentation more than building-to-building or campus-to-campus communication. That's changing; companies now have (or are coming out with) Packet-Over-SONET and Wave Division Multiplexing (WDM) interfaces to put LAN traffic to the (really fast)WAN. They're expensive like crazy, but companies that use 'em are getting LAN performance (well, really close to LAN performance) from coast-to-coast.

So, there ya go.....

FWIW

Scott
 


<< The &quot;Switch&quot; part of the consumer &quot;Switch-Router&quot; applies only to the Ethernet ports (all layer two stuff). When the packet is passed to the router part of the device, everything is happening like a traditional router: the packet is examined to see where it's going, where it came from, special handling (changing the address for NAT, etc). That process of examination and forwarding is done in software, and takes a fair amount of time. >>



so does that apply to a regular cable/dsl router? or is that for the more expensive switch-router?
 
Does not apply to consumer-grade stuff.

L3 Switching costs lots o' bucks.

The &quot;switching&quot; stuff in consumer-grade products only applies to the inside ports.

FWIW

Scott
 

so what's the advantage of using a layer-3 switch then? only segmentation? how does this differ from a router connected to a switch?

aren't backbone routers usually switch routers?
 
A layer three switch is &quot;much&quot; faster than most routers, and usually has more ports for less money. Getting 24 Fast Ethernet ports into a router would be ugly, extremely expensive, and would require a killer processing engine. L3 switches also operate at &quot;wire speed,&quot; and most routers do not. L3 switches ain't cheap, but compared to a router rigged to accomodate the same functionality, it's a definate cost savings (with better performance).

Since L3switches can operate at L2 as well, you can set up groups of ports that route to other groups of ports...virtually the same as having a switch hanging off a router...but the L3switch is faster in getting the data through. A common application is to use the L3 switch to aggregate L2 switches in-use at the access level....but you should verify that the L3 switch has the necessary MAC-per-port capacity.

For the hard-core network administrator, that has to deal with end-users adding extra networking components (hubs under their desk, for example), an L3 switch would allow the admin to set 30 bit subnet masks to each end-users subnet (one subnet per user, each subnet having only one valid address for the end-user connection)(Doug/L3Guy once recommended this for our call center's network...you know them engineer types...always messing with the LAN).

There are a number of reasons to use L3 switches, but most of the reasons stem from the segmentation functionality.

AND (sorry for singing an old song here....) a poorly implemented L3 switch does way more damage to performance and administrative overhead than it does good. Like any major infrastructure change, it MUST be well-planned before the hardware ever hits the network.

One of the major considerations is when a WAN is immediately adjacent to a highly segmented network (especially if it's using RIP as a routing portocol). If the LAN has tens/dozens/hundreds of routed segments, and you fail to properly manage the routing protocol updates, then you're sending a great deal of (usually useless) traffic over the WAN and wasting a lot of bandwidth.

Bottom line: an L3switch IS a router, a really fast router, most of the time you can use it the same as you would a &quot;traditional&quot; router (there are exceptions, of course). The main role of an L3switch is segmentation, and all the good/bad things that segmentation brings with it. Routers are usually still desirable/necessary despending on the network design and policies(usually WAN access).

FWIW

Scott
 


<< [awesome technical details snipped]The main role of an L3switch is segmentation, and all the good/bad things that segmentation brings with it. Routers are usually still desirable/necessary despending on the network design and policies(usually WAN access). >>

The designer's mantra for this--&quot;Switch where you can, route where you must&quot;--is the general rule for the price/performance reasons ScottMac talked about. And he's SO right...the importance of planning before implementing L3 switching (or any other kind of intelligent network device) cannot be stressed enough.

Case in point: I once met this guy who had been &quot;promoted&quot; to head of systems for a multi-state cardiology practice group. He was obviously out of his league in the position, since he had never been involved in networking before. While we were chatting, he dragged out a topology diagram (left by the old system designer) and asked me, &quot;Say, if you were going to put a router in here, where would it go?&quot; I said, &quot;Why? What are you guys planning to do?&quot; He said, &quot;Well, someone in our group talked to one of our vendors and they said we should remove all our routers, stick in some switches, and automatically be running orders of magnitude faster. But then another vendor said we should implement some faster routers instead of switches. Where would you put them?&quot; I looked him straight in the eye for a minute, then shook my head, and said, &quot;Nice talking to you. Good luck.&quot;
 
LOL!
I did recommend a 30 bit subnet mask solution for a problem network.
It would have helped, too! However, understand it was a RADICAL solution
that I would use only as a last resort.

Since each port has 2 valid addresses in a /30, and one is used by the L3 switch,
it is not possible to have ip address conflicts. The engineers had a habit of adding
&quot;extra&quot; devices. Their method of locating a usable address was pinging until
something did not answer. Special. For those of you that don't work with this
every day, its important to document ip addresses in any network above 3 or 4 PC's.

The other problem that this would fix was blocking all routing protocols originated
from the &quot;extra&quot; devices. Since all IGP (Interior Gateway Protocol = Corporate)
routing protocols are broadcast or multicast, the L3 switch would block them unless
it was configured to listen. There are other ways of dealing with this, such as
authentication or passive interfaces, but effective. The broadcast blocking also
made sniffing much less useful on the net, because it blocked everything besides
unknown unicasts that were flooded and traffic directed to that port.

Sorry for the ramble;

Doug

 
Back
Top