Login | Register For Free | Help
Search for: (Advanced)

Mailing List Archive: Cisco: NSP

Maximum switched ethernet daisy chain size

 

 

Cisco nsp RSS feed   Index | Next | Previous | View Threaded


pavel.skovajsa at gmail

Jun 7, 2008, 8:50 AM

Post #1 of 4 (5403 views)
Permalink
Maximum switched ethernet daisy chain size

Hello all,
Recently I have found an interesting issue.
I have found that on our network the packets that traverse more then
17 (18 and more) L2 ethernet switching hops all connected with MM
fiber 1G links have lowered probability of arrival.
The way I have found this is that our big ethernet ring with 30+
switches was cut (fiber issue) behind 18th switch from L3 switch, and
everybody hosted after 18th switch has some interesting packet loss
issues.See the smokeping details for 18th switch on
http://img171.imageshack.us/img171/3777/hop18kn4.jpg.

This is not very logical for me, as switched ethernet works based on
store&forward method so no old-school 5-4-3 rule applies as in shared
collision (hub) environment. Anybody has a good explanation why adding
more L2 hops to the path tends to kill packets?

Regards,
Pavel

p.s. yes I know that building large ethernet rings is generally not a
good idea, however sometimes due to large distances you have no other
choice
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


r.engehausen at gmail

Jun 7, 2008, 10:24 AM

Post #2 of 4 (5332 views)
Permalink
Re: Maximum switched ethernet daisy chain size [In reply to]

I suspect you are having some bandwidth issues. Switches only have so
much buffer space and momentary high usage can cause an overflow and
packet loss

As an example I did a little googling. I found a Cisco article on GE
adapters for the 65xx series that shows the various queue sizes for
different blades.

http://www.cisco.com/warp/public/cc/pd/si/casi/ca6000/prodlit/c60ge_ds.htm

Roy

Pavel Skovajsa wrote:
> Hello all,
> Recently I have found an interesting issue.
> I have found that on our network the packets that traverse more then
> 17 (18 and more) L2 ethernet switching hops all connected with MM
> fiber 1G links have lowered probability of arrival.
> The way I have found this is that our big ethernet ring with 30+
> switches was cut (fiber issue) behind 18th switch from L3 switch, and
> everybody hosted after 18th switch has some interesting packet loss
> issues.See the smokeping details for 18th switch on
> http://img171.imageshack.us/img171/3777/hop18kn4.jpg.
>
> This is not very logical for me, as switched ethernet works based on
> store&forward method so no old-school 5-4-3 rule applies as in shared
> collision (hub) environment. Anybody has a good explanation why adding
> more L2 hops to the path tends to kill packets?
>
> Regards,
> Pavel
>
> p.s. yes I know that building large ethernet rings is generally not a
> good idea, however sometimes due to large distances you have no other
> choice
> _______________________________________________
> cisco-nsp mailing list cisco-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>
>

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


clinton at scripty

Jun 7, 2008, 11:22 AM

Post #3 of 4 (5330 views)
Permalink
Re: Maximum switched ethernet daisy chain size [In reply to]

Are you running spanning-tree on the large metro Ethernet ring? Default
STP settings limit you to 20 switches from the STP root for a stable
topology. You have to consider the worst case distance from the STP
root rather than the steady state one.


Pavel Skovajsa wrote:
> Hello all,
> Recently I have found an interesting issue.
> I have found that on our network the packets that traverse more then
> 17 (18 and more) L2 ethernet switching hops all connected with MM
> fiber 1G links have lowered probability of arrival.
> The way I have found this is that our big ethernet ring with 30+
> switches was cut (fiber issue) behind 18th switch from L3 switch, and
> everybody hosted after 18th switch has some interesting packet loss
> issues.See the smokeping details for 18th switch on
> http://img171.imageshack.us/img171/3777/hop18kn4.jpg.
>
>
>


--
=======================================================
Clinton Work
Airdrie, AB

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


mack at exchange

Jun 7, 2008, 11:08 PM

Post #4 of 4 (5308 views)
Permalink
Maximum switched ethernet daisy chain size [In reply to]

> ------------------------------
>
> Message: 3
> Date: Sat, 7 Jun 2008 17:50:59 +0200
> From: "Pavel Skovajsa" <pavel.skovajsa [at] gmail>
> Subject: [c-nsp] Maximum switched ethernet daisy chain size
> To: cisco-nsp [at] puck
> Message-ID:
> <323aca890806070850me92fb8bq53e7d8ef4e0d2841 [at] mail>
> Content-Type: text/plain; charset=ISO-8859-1
>
> Hello all,
> Recently I have found an interesting issue.
> I have found that on our network the packets that traverse more then
> 17 (18 and more) L2 ethernet switching hops all connected with MM
> fiber 1G links have lowered probability of arrival.
> The way I have found this is that our big ethernet ring with 30+
> switches was cut (fiber issue) behind 18th switch from L3 switch, and
> everybody hosted after 18th switch has some interesting packet loss
> issues.See the smokeping details for 18th switch on
> http://img171.imageshack.us/img171/3777/hop18kn4.jpg.
>
> This is not very logical for me, as switched ethernet works based on
> store&forward method so no old-school 5-4-3 rule applies as in shared
> collision (hub) environment. Anybody has a good explanation why adding
> more L2 hops to the path tends to kill packets?
>
> Regards,
> Pavel
>
> p.s. yes I know that building large ethernet rings is generally not a
> good idea, however sometimes due to large distances you have no other
> choice
>
>

This sounds like a spanning-tree diameter problem.
Spanning-tree limits the number of intermediate switches between the root
and each end-point. Without tuning this value is 7.

This link describes possible issues with diameter:
http://www.cisco.com/warp/public/473/16.html#awkward

This link may help you resolve the problem:
http://www.cisco.com/en/US/tech/tk389/tk621/technologies_tech_note09186a0080094954.shtml

--
LR Mack McBride
Network Administrator
Alpha Red, Inc.

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

Cisco nsp RSS feed   Index | Next | Previous | View Threaded
 
 


Interested in having your list archived? Contact Gossamer Threads
 
  Web Applications & Managed Hosting Powered by Gossamer Threads Inc.