From: Mike Williams (ccie2be@swbell.net)
Date: Wed Apr 23 2003 - 22:26:18 GMT-3
Careful here. Remember there are A LOT of different Cat switches and
many of them have different architectures. The 3550 has a shared memory
architecture, so (if my understanding of the technology is correct) when
a frame comes in, it's stored in shared memory, and then only the port
(or ports if spanning) that it will exit are passed a pointer to that
frame in memory. Very little overhead there. However, I'm not sure how
long it keeps that frame in that same memory location, so if the span
port is current outputing another frame, I can't say that it would keep
that frame in memory, or copy it to a buffer. That would become an ugly
scenario as it's possible that your source ports (if you're spanning
more than one port as the input) could burst to an excessively high
speed that would overrun the output buffer of the output span port. I'm
sure at this point Cisco engineers (the people that build this hardware)
are really the only ones that truly know what kind of performance hit a
switch might take in this situation trying to juggle these spanned
frames, etc...
But all in all I would agree with the original post that there is some
sort of ASIC or something that handles the RSPAN technique to keep the
load off the CPU.
2 cents.
Mike W.
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Chris Home
Sent: Wednesday, April 23, 2003 5:53 PM
To: David Buechner; ccielab@groupstudy.com
Subject: Re: 3550 - RSPAN - Reflector Port
In fact when a cat switch receives ANY frame it is copied to every port.
When a switch receives a frame, as soon as it gets access to the
backplane the frame is copied to every port as it travels. When a
forwarding decision is made the interfaces that should not forward are
told to flush their buffer of the frame or ignore it (I am not exactly
sure which. I have heard both ignore and flush). The forwarding ports
are left to forward. So a span port simply never gets told to flush or
ignore any packets. Little to no performance loss. There is also an
excellent document on this and the 6500 architecture that covers more of
this kind of stuff. Very informative and interesting doc.
I am not sure how this applies on dCEF cards. Since the forwarding
decision is made at the module. I think it might be the same, the
forwarding decision is just made sooner. I dunno really.
----- Original Message -----
From: "David Buechner" <dbuechn@attglobal.net>
To: <ccielab@groupstudy.com>
Sent: Wednesday, April 23, 2003 2:53 PM
Subject: RE: 3550 - RSPAN - Reflector Port
> Daniel,
>
> Here's how I understand it:
>
> First, let's start with a SPAN session. In SPAN you have some traffic
that
> you're monitoring which is then duplicated on a port local to the
> switch
so
> that it is visible to a locally attached device (such as a protocol
> analyzer). One of the things that I suspect is important to switch
> performance here is that very little processing is done on the
> duplicated packet - it's just copied to the monitor port.
>
> Now consider the RSPAN scenario. Here, instead of copying the frame
> to a particular port you need to get it into a particular VLAN and
> then switch it as new traffic for that VLAN, thus enabling delivery to
> a remote destination port. The mechanism on the local switch to do
> this is the reflector port. In essence the reflector port acts
> similarly to a destination port, only instead of transmitting the
> packet out the physical port it transmits it out the RSPAN VLAN. I
> don't know the hardware intimately, but I would guess that maybe the
> ASIC in the port can handle "reflecting" the traffic back to the RSPAN
> VLAN thereby keeping the switch CPU from having to modify the packet.
> (I'd be happy to hear from someone who knows the hardware better to
> see if my WAG is right! :-))
>
> If I'm wrong in any of this I'd very much appreciate commentary!
>
> David
>
> At 07:25 am 4/23/2003 -0600, groupstudy wrote:
>
>http://www.cisco.com/en/US/products/hw/switches/ps646/products_configur
>atio
n_guide_chapter09186a00800c6f4c.html
> >Pretty good explanation
> >
> >-----Original Message-----
> >From: Daniel Cisco Group Study [mailto:danielcgs@imc.net.au]
> >Sent: Wednesday, April 23, 2003 6:19 AM
> >To: ccielab@groupstudy.com
> >Subject: 3550 - RSPAN - Reflector Port
> >
> >
> >Hi all,
> >
> >Can someone explain the use / purpose of the reflector port when
> >setting up RSPAN?
> >
> >The docs are a bit vague on this.
> >
> >Thanks in advance.
> >
> >Daniel
> >
> >
> >
> >*********************************************************************
> >*
> >This email and any files transmitted with it are confidential and
> >intended solely for the use of the individual or entity to whom they
> >are addressed. If you have received this email in error please notify
> >the system manager.
> >This footnote also confirms that this email message has been swept by
> >MIMEsweeper for the presence of computer viruses.
> >www.mimesweeper.com
>
>**********************************************************************
This archive was generated by hypermail 2.1.4 : Thu May 01 2003 - 13:36:04 GMT-3