Velocity Reviews - Computer Hardware Reviews

Velocity Reviews > Newsgroups > Computing > Cisco > Re: LAN<-->WAN<-->LAN ; L3 switches or Routers??

Reply
Thread Tools

Re: LAN<-->WAN<-->LAN ; L3 switches or Routers??

 
 
thrill5
Guest
Posts: n/a
 
      07-20-2006
This is not entirely true. A "L3 switch" does not do QoS. Yes a "L3 Switch"
can do QoS marking, but it cannot do policing, queuing (at layer 3) WRED,
etc. A L3 switch does CoS (class of service), which is limited by hardware
(number or receive queues, output queues and priority queues) on each port.
FLEXWAN and OSM cards for the 6500 series are boards that let you do the
same QoS stuff that you can do on router interface. Its also why they are
big bucks.

Scott
"stephen" <(E-Mail Removed)> wrote in message
news:cZctg.1221$(E-Mail Removed)...
> "Merv" <(E-Mail Removed)> wrote in message
> news:(E-Mail Removed) oups.com...
>>
>> > So something to consider is what hardware you need to interface to the

> WAN.
>>
>> This is a good point as while Cisco high-end switches like the 6500
>> have FLEXWAN card to support more than just Ethernet connectivity this
>> is not the case on their low end switches.

>
> this is a good point to mention to the OP that the difference between
> routers and L3 switches these days is more about marketing and the "bias"
> in
> the box design than engineering.
>
> originally - routers were basically software, and switches were hardware.
> now boxes with reasonable performance are usually a "blend" somewhere
> between those 2 extremes.
>
> the rule of thumb is that if all the interfaces are "lan like" - then it
> probably gets called a switch, and if you are using older style WAN
> interfaces such as T1 / E1, Frame Relay or ATM it probably is called a
> router - but as other have said there are exceptions to both of these.
>>
>> So you might need routers and switches depending on the WAN transport
>> options available.
>>

> --
> Regards
>
> http://www.velocityreviews.com/forums/(E-Mail Removed) - replace xyz with ntl
>
>



 
Reply With Quote
 
 
 
 
stephen
Guest
Posts: n/a
 
      07-21-2006
"thrill5" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed)...
> This is not entirely true. A "L3 switch" does not do QoS.


as with most computer things "it depends" - it doesnt have to, but any high
end network equipment these days that doesnt "do" QoS doesnt get bought -
QoS features are mandatory when a company generates a tick list for a big
procurement, and the manufacturers want their kit to get bought.

Yes a "L3 Switch"
> can do QoS marking, but it cannot do policing, queuing (at layer 3) WRED,
> etc. A L3 switch does CoS (class of service), which is limited by

hardware
> (number or receive queues, output queues and priority queues) on each

port.

FWIW i just did a lot of lab testing on a Cat6, Sup 720-3B, 6724-SFPs.

policing, Q managment (2 Qs inbound, 4 out), and WRR (which is sort of
modified WRED) are all there.

it is the same rich set of features as you get on a software router, but it
did everything we wanted, apart from shaping to fractional Gig E while also
doing QoS queue management.

> FLEXWAN and OSM cards for the 6500 series are boards that let you do the
> same QoS stuff that you can do on router interface. Its also why they

are
> big bucks.


Agreed.

in our case it was cheaper to use more bandwidth and do GigE "wire speed"
across a WAN than to have the richer features needed to handle lower speed
links.

>
> Scott
> "stephen" <(E-Mail Removed)> wrote in message
> news:cZctg.1221$(E-Mail Removed)...
> > "Merv" <(E-Mail Removed)> wrote in message
> > news:(E-Mail Removed) oups.com...
> >>
> >> > So something to consider is what hardware you need to interface to

the
> > WAN.
> >>
> >> This is a good point as while Cisco high-end switches like the 6500
> >> have FLEXWAN card to support more than just Ethernet connectivity this
> >> is not the case on their low end switches.

> >
> > this is a good point to mention to the OP that the difference between
> > routers and L3 switches these days is more about marketing and the

"bias"
> > in
> > the box design than engineering.
> >
> > originally - routers were basically software, and switches were

hardware.
> > now boxes with reasonable performance are usually a "blend" somewhere
> > between those 2 extremes.
> >
> > the rule of thumb is that if all the interfaces are "lan like" - then it
> > probably gets called a switch, and if you are using older style WAN
> > interfaces such as T1 / E1, Frame Relay or ATM it probably is called a
> > router - but as other have said there are exceptions to both of these.
> >>
> >> So you might need routers and switches depending on the WAN transport
> >> options available.
> >>

> > --
> > Regards
> >
> > (E-Mail Removed) - replace xyz with ntl
> >

--
Regards

(E-Mail Removed) - replace xyz with ntl


 
Reply With Quote
 
 
 
 
J
Guest
Posts: n/a
 
      07-21-2006
stephen wrote:
> FWIW i just did a lot of lab testing on a Cat6, Sup 720-3B, 6724-SFPs.
>
> policing, Q managment (2 Qs inbound, 4 out), and WRR (which is sort of
> modified WRED) are all there.
>
> it is the same rich set of features as you get on a software router, but it
> did everything we wanted, apart from shaping to fractional Gig E while also
> doing QoS queue management.


Comparing a 6500 to a just a router or just a switch is not a fair
comparison. A 6500 contains both a router and a switch (plurality in
some cases). There is no apples to apples comparison in a chassis that
has MLS capabilities. If you want to compare a basic router and a
basic switch then compare a 3825 without any EtherSwitch modules and a
3750-EMI. That's a fair comparison between a router and a L3-switch.

> > FLEXWAN and OSM cards for the 6500 series are boards that let you do the
> > same QoS stuff that you can do on router interface. Its also why they

> are
> > big bucks.

>
> Agreed.
>
> in our case it was cheaper to use more bandwidth and do GigE "wire speed"
> across a WAN than to have the richer features needed to handle lower speed
> links.


This makes good sense. I'm also opposed to blending network layers in
large chassis. You could run an entire ISP out of a single 7600 but
that would involve blending core, distribution, access, and border
router functions (I always separate border routers into a 4th network
layer) into a single chassis. That breaks most of the design
principals of modern day LAN/WAN design. Just because the device
you're working with can do routing, it doesn't make it a good router.
Likewise for interfaces; just because you can put 16 FastEthernet
interfaces in a router doesn't mean it makes a good switch. This also
applies to circuits and their purposes. Just because you can get an
Ethernet hand-off for your Internet circuit doesn't mean you should
forego a real router and plug it into a L3 switch.

One place that a WAN Ethernet hand-off does allow you to fudge on is
firewalls. I don't have any problem at all with eliminating the border
router for a small non-multi-homed office and replacing it with a
purely-Ethernet firewall. There are design applications that may
require a connection outside of the firewall but most of our customers
don't have these requirements. If the hand-off had been a couple T1s
they would have been stuck with a border router.



J

 
Reply With Quote
 
stephen
Guest
Posts: n/a
 
      07-22-2006
"J" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed) oups.com...
> stephen wrote:
> > FWIW i just did a lot of lab testing on a Cat6, Sup 720-3B, 6724-SFPs.
> >
> > policing, Q managment (2 Qs inbound, 4 out), and WRR (which is sort of
> > modified WRED) are all there.
> >
> > it is the same rich set of features as you get on a software router, but

it
> > did everything we wanted, apart from shaping to fractional Gig E while

also
> > doing QoS queue management.

>
> Comparing a 6500 to a just a router or just a switch is not a fair
> comparison. A 6500 contains both a router and a switch (plurality in
> some cases). There is no apples to apples comparison in a chassis that
> has MLS capabilities. If you want to compare a basic router and a
> basic switch then compare a 3825 without any EtherSwitch modules and a
> 3750-EMI. That's a fair comparison between a router and a L3-switch.


you are tlkaing about hybrid mode where you use CatOS for the L2 aspects of
the switch and IOS on the MSFC to provide an embedded router. but you dont
have to run a Cat6k like that - and in some configs you cant.

when you use the DFC blade you run the Cat6k as an IOS only switch - it
basically follows the same design as a 3750, just extra scale and some more
flexibility.
>
> > > FLEXWAN and OSM cards for the 6500 series are boards that let you do

the
> > > same QoS stuff that you can do on router interface. Its also why

they
> > are
> > > big bucks.

> >
> > Agreed.
> >
> > in our case it was cheaper to use more bandwidth and do GigE "wire

speed"
> > across a WAN than to have the richer features needed to handle lower

speed
> > links.

>
> This makes good sense. I'm also opposed to blending network layers in
> large chassis. You could run an entire ISP out of a single 7600 but
> that would involve blending core, distribution, access, and border
> router functions (I always separate border routers into a 4th network
> layer) into a single chassis.


FWIW a 7600 is Cat6k, AFAIR just different chassis, vertical mounted blades
and restricted to IOS only mode.

That breaks most of the design
> principals of modern day LAN/WAN design. Just because the device
> you're working with can do routing, it doesn't make it a good router.
> Likewise for interfaces; just because you can put 16 FastEthernet
> interfaces in a router doesn't mean it makes a good switch. This also
> applies to circuits and their purposes. Just because you can get an
> Ethernet hand-off for your Internet circuit doesn't mean you should
> forego a real router and plug it into a L3 switch.


Actually we are delivering Internet access to customers at work - and often
that is exactly what they do.

For 10 and 100 M, it doesnt matter that much whether you have a router or a
switch, since a mid range router like a 3845 is fast enough.

A "real" router with the horsepower to terminate a GigE WAN link is an
expensive toy - A stackable Cat has good enough forwarding to substitute
(although there isnt as much software processing available, memory for
routing tables etc) at a small fraction of the cost.

>
> One place that a WAN Ethernet hand-off does allow you to fudge on is
> firewalls. I don't have any problem at all with eliminating the border
> router for a small non-multi-homed office and replacing it with a
> purely-Ethernet firewall. There are design applications that may
> require a connection outside of the firewall but most of our customers
> don't have these requirements. If the hand-off had been a couple T1s
> they would have been stuck with a border router.
>
>
>
> J

--
Regards

(E-Mail Removed) - replace xyz with ntl


 
Reply With Quote
 
 
 
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Cisco Switches vs HP ProCurve Switches Greg Cisco 5 07-01-2013 01:48 AM
switches, spanning tree question regarding designated ports and switches alefveld@versatel.nl Cisco 1 12-30-2008 01:24 AM
KVM Switches and Wireless Devices =?Utf-8?B?RF9JZ2Vs?= Wireless Networking 1 09-27-2004 12:30 AM
Mozilla command line switches William W. Plummer Firefox 1 08-18-2004 03:37 PM
native vlan mismatch on 2 2924 switches w/ only 1 vlan defined (same on both switches) avraham shir-el Cisco 4 07-20-2004 08:08 AM



Advertisments