[ednog] Anyone deploying MSTP and/or RSTP as alternatives to spanning tree?

David Farmer farmer at umn.edu
Tue Apr 26 19:19:37 EDT 2005


MSTP is 802.1s and it uses RSTP 802.1w as the basis for converging 
on a Tree rather than the older 802.1d algorithm.

The following is the design Guide we used to help implement.

http://www.cisco.com/application/pdf/en/us/guest/tech/tk621/c1501/ccmigration_09186a0080174993.pdf

This doc has good transition information in it.

Something about how we are using it.  We have 17 Core Nodes 
connected by an L3 Core Network, each of our Core Nodes are a 
MSTP Region and is the root for its Region.  Each Region has a single 
MSTP Instance responsible for all VLANs.

We generally do not have L2 loops in our network but use MST to 
protect us from accidental loops.

The exception to this is our Data Center infrastructure, which will have 
a separate MSTP Region with multiple Instances, probably only 2, 
supporting different VLANs.   The two Data Center Core Nodes will be 
doing HSRP.  The Data Center Core Node which is the Primary for 
HSRP will be the Spanning Tree Root for the MSTP Instance 
containing those VLANs for which it is the HSRP primary router. 

We have had a generally good experience with MSTP.

We had one small problem.  We configure new switchs via a script, the 
old script turned on VTP before MST.  This wasn't a problem until we 
got to our largest Core Node that has 178 VLANs.  So, 3750s support 
128 spanning tree instances and default to PVST (Per VLAN Spanning 
Tree) as do most Cisco Switches.  So when VTP was enabled it tried 
to create 178 spanning trees, it could so it disable Spanning Tree for 
about 50 VLANs, and they remain disable after MST was configured. 

And then an accidental loop got created during an install, and BAM!!!  
Actually, this happened many time before during install, crossed fibers, 
etc...  But this time we didn't have spanning tree on a bunch of VLANs.  

So first we killed the loop, then we re-enabled spanning tree on all the 
VLANs, and finally changed the order of the script so now it enables 
MST first then enables VTP.  

This really wasn't MST's fault, mostly a combination of a bad script on 
our part and a nasty choice by an IOS programmer to leave spanning 
tree disabled on some VLANs when MST gets enabled.  But man it 
hurt!!!

On 26 Apr 2005 Laura Kristoff wrote:

> Cisco has something called MSTP that is supposedly their
> interpretation of IEEE 802.1s.  I'm still trying to understand how it
> works but they say it uses RSTP (802.1w) which allows for faster
> convergence because it uses an explicit handshake to figure out the
> forwarding path.  I think the MSTP part cuts down on the amount of
> BPDU traffic and allows for load balancing as well as redundancy.
> 
> Before delving into this any deeper, I was wondering if anyone has
> considered this as an alternative to running STP on their switches? 
> We try to limit the number of vlans that we trunk around campus, but
> there's always some amount of trunking we end up having to do.
> 
> Does anyone have any experience (good or bad) with these technologies?
> 
> 
> 
> --
> Laura Kristoff
> IT Telecommunications and Network Services
> 847-491-4094   l-grill at northwestern.edu
> 
> 
> _______________________________________________
> ednog mailing list
> ednog at puck.nether.net
> https://puck.nether.net/mailman/listinfo/ednog
> 



=================================================
David Farmer				Email:	farmer at umn.edu
Office of Information Technology
University of Minnesota			Phone:	612-626-0815
2218 University Ave SE			Cell:	612-812-9952
Minneapolis, MN 55414-3029		FAX:	612-624-4035
=================================================



More information about the ednog mailing list