At 02:51 AM 5/22/02, Chuck wrote: >sorry to keep harping on this one, but I'm actually learning something here. >Besides, my big project at work these days is working with a large >university, replacing their campus physical and switch infrastructure. I'm >finding the this discussion fascinating for that reason as well.
Well, it might not have any real-world relevance. ;-) >If I read my source correctly, the max age field is supposed to be 2 bytes, >and is supposed to be a time value, That sounds like Message Age actually. Message Age times the age of a BPDU. The root sends a BPDU with Message Age set to zero. Each bridge adds 1. So it is sort of a hop count. In a functioning network, the bridges don't pay much attention to this since BPDUs are refreshed every 2 seconds. >with the min being 1/256 second and the >max being 256 seconds. other than in the initial STP process ( or >recalculation ) In a non-functioning network, the Maximum Age threshold comes into play. Its default is 20. You can change it (at the root bridge only; the others learn it from the root). The Maximum Age controls the size of the network, but it also has a much more important purpose, which is to start reconvergence. As I understand it, the BPDU arrives with the Message Age = to hop count. But the BPDU continues to age until it reaches Maximum Age. If the Root Bridge fails, another bridge will notice the Message Age reach the Maximum Age and start the process of taking over as the Root Bridge. If the Root Bridge doesn't fail, but a path to the Root Bridge fails, if an alternate path exists, a blocking port on a downstream bridge transitions to listening, learning, and forwarding after it notices Message Age reach Maximum Age. If a Root Port fails, another port on the bridge where the failure occurred may transition directly into the listening and learning states without waiting for Maximum Age. It's horridly complex. ;-) >the BPDU would for all practical purposes be time from the >root. Correct? My source tells me only the fields and their values, and >nothing about functionality. It would appear that the max age field tells >the local switch how old a message can be before it is disregarded, or >causes some other action to be taken. The message age field is the actual >age as per the process you describe below - incremented by each bridge along >the way. Yes, that sounds right. >The root path cost is used to advertise how far this bridge is from the >root? "hops"? No, cost like in OSPF. Each interface has a cost: Link Speed Recommended Cost Value 4 Mbps 250 10 Mbps 100 16 Mbps 62 100 Mbps 19 1 Gbps 4 10 Gbps 2 >counting on my fingers, a max distance of 20 from the root is a whole lot >different than a max diameter of 7. The 7 is a recommended value. Try even finding it in Radia Perlman's book!? ;-) I hope I didn't just confuse matters even more. In addition to the Perlman bible, try the Clark and Hamilton holy writ. Priscilla >Chuck > > >""Priscilla Oppenheimer"" wrote in message >[EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > > There's nothing in the STP frames to enforce a 7 hop diameter. But there >is > > the Message Age field in the BPDUs. Each bridge (switch) adds one to the > > Message Age when the switch propagates the BPDU downstream. > > > > The Maximum Age threshold is 20. If a BPDU gets to a switch with the > > Message Age already at 20, it will think that the tree needs reconverging. > > This would get ugly if switches on the edges were always trying to > > reconverge. So, the max size from that viewpoint is 20 from the root. > > > > But 7? I really think DEC threw that in as a precaution. It's interesting > > that IBM was saying the same thing about source route bridging at the time > > (max bridges is 7). (But try finding 7 in IEEE 802.1D Annex C, the >official > > standardization of source-route transparent bridging. The RIF can actually > > hold info for 14 rings and 13 bridges.) > > > > Back to the real subject at hand, the 7 max for STP is mentioned as a > > recommended value in Table 8.2 Maximum Bridge Diameter of IEEE 802.1D and > > is defined as "The maximum number of Bridges between any two points of > > attachment of end stations." > > > > Then it's discussed again in Appendix B "B.3.1.2 Basis of choice." This > > section is pretty incomprehensible, but, as far as I can tell, the main > > reason for the choice of parameters is to minimize the lifetime of a data > > (user) frame travelling across the switched network. > > > > Regarding gigastack, it sounds like the answer that Steven got from Cisco > > is that each switch counts as a hop, so if STP is enabled, each counts > > toward the _theorectical_ 7 hop count limit. > > > > But I bet you're right also that STP could be disabled with gigastack. It > > sounds like the topology is already a single linear branch (stack) with no > > loops. There's no need to prune it into a tree. But I'm way out on a limb > > now. ;-) > > > > Priscilla > > > > At 06:34 PM 5/19/02, Chuck wrote: > > >you know, it suddenly occurs to me that I have been barking up the wrong > > >[spanning] tree, so to speak. > > > > > >Let me guess - there is no maximum STP diameter in actuality because >there > > >is mechanism for enforcing a max diameter. The BPDU's apparently contain >a > > >field which shows distance from the root, and this value is incremented >each > > >time it crosses a bridge. If that field is the root path cost field, then > > >this is a four byte value and that means a spanning tree could >theoretically > > >have a max distance from the root of 64000 or so? > > > > > >It's just that the recommendation in terms of "best practice" is diameter >of > > >7. > > > > > >thanks to Marty A. for providing the link that was the spike that finally > > >began to sink through this thick head. > > > > > > > > >""Chuck"" wrote in message > > >[EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > > > > STP is really not an issue in the kind of application where gigastack > > >makes > > > > sense. For example, take an office of 400 users plus servers and > > printers, > > > > occupying a contiguous space. Basic file and print sharing plus an > > >internet > > > > connection. Rather than buy a honking 65xx, you throw in a few >3550-48's > > >and > > > > gigastack them. The electronics work in conjunction with the switch OS >to > > > > create a half duplex bus between the switches. ( The interesting thing >is > > > > that electronics are apparently smart enough to determine if there are > > >only > > > > two devices stacked, in which case the bus is full duplex. ) > > > > > > > > That's the question about gigastack - whether the entire stack is >treated > > >as > > > > one switch, the way it is for management purposes, or if standard STP > > > > applies. We had a thread on this a few weeks ago, but none of us could > > >find > > > > an answer in the Cisco documentation. that's why I asked Steven ( who > > >asked > > > > Cisco ) what Cisco had to say about spanning tree over a gigastack >setup. > > > > I'm willing to bet that in a gigastack situation, that STP is disabled >( > > > > where it would be enabled if you stacked the switches using the 10/100 > > > > ports, or using regular gbics and daisy chaining the gig ports. ) But >I > > > > can't find documentation that clarifies. > > > > > > > > This does give me some food for thought. I'm working with a very large > > > > university on a switch design. I am pretty sure that they have several > > > > instances where their diameter is greater than seven ( building to > > >building > > > > to building to core ). However, being the sophisticates they are, I am > > > > certain they are running spanning tree per vlan, and in that case the > > tree > > > > would go back only to the core, and so would have a diameter no >greater > > >than > > > > 4 in any place I can see from their setup. > > > > > > > > this seems to be another one of the great discussions that make this > > group > > > > valuable. > > > > > > > > > > > > > > > > ""Priscilla Oppenheimer"" wrote in message > > > > [EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > > > > > At 03:51 PM 5/19/02, Chuck wrote: > > > > > >were the people you asked able to address the max STP diameter of 7 > > >issue > > > > > >then? > > > > > > > > > > Maybe they blew off the max STP diameter of 7 issue. ;-) It is just >a > > > > > recommendation. Even IEEE 802.1D just uses that word >(recommendation). > > > > > There's nothing in the protocol that would stop you from having a > > larger > > > > > network, is there? The Max Age timer defaults to 20 seconds, so that > > > > > doesn't limit you to 7 hops. > > > > > > > > > > Also, with gigaswitch, do all those 9 switches remain in the >spanning > > > > tree? > > > > > I don't know anything about gigaswitch, as you can probably tell, >but I > > > > > would think the 7 limit would only apply to the logical topology, if >it > > > > > applies at all. > > > > > > > > > > Priscilla > > > > > > > > > > > > > > > > > > > > >""Steven A. Ridder"" wrote in message > > > > > >[EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > > > > > > > I asked Cisco, and STP treats the stack as separate switches. > > > > > > > > > > > > > > > > > > > > > ""Chuck"" wrote in message > > > > > > > [EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > > > > > > > > in your travels have you ever run into the gigastack, and its > > > > relation > > > > > >to > > > > > > > > STP? > > > > > > > > > > > > > > > > the question came up a short time ago. The Cisco gigastack > > > > > documentation > > > > > > > > does not cover STP - only how gigastack works. > > > > > > > > > > > > > > > > essentially, using daisy chained gigastack connections, one >can > > > > connect > > > > > >up > > > > > > > > to 9 switches. If I understand correctly, the Cisco >electronics > > >and > > > > > >switch > > > > > > > > OS consider this stack a single entity for management >purposes - > > > > i.e. > > > > > >can > > > > > > > be > > > > > > > > managed from a single IP address. However, Cisco does not say >one > > > > way > > > > > or > > > > > > > > another if the electronics and the switch OS treat this stack >as > > a > > > > > >single > > > > > > > > device for STP purposes. > > > > > > > > > > > > > > > > I'm guessing, based on this and other discussions that it >would > > >have > > > > to > > > > > >be > > > > > > > a > > > > > > > > single device for STP purposes. It just would be nice if Cisco > > >would > > > > > > > provide > > > > > > > > the specific information. > > > > > > > > > > > > > > > > That brings up the corollary question - if you have on two > > >switches > > > > > > > > gigastacked, then how does STP come into play. In a >configuration > > > > such > > > > > >as > > > > > > > > this, again if I understand the documentation correctly, the > > > > > electronics > > > > > > > and > > > > > > > > the switch OS behave differently. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > ""Leigh Anne Chisholm"" wrote in message > > > > > > > > [EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > > > > > > > > > Actually, the 5-4-3 rule has everything to do with detecting > > > > > >collisions. > > > > > > > > > It's > > > > > > > > > a limiting factor of distance so that a collision will be > > >detected > > > > > > > within > > > > > > > > the > > > > > > > > > first 64 bytes of a frame's transmission (also known as > > >Ethernet's > > > > > > > minimum > > > > > > > > > frame size). It's preferable to detect a collision before >the > > > > frame > > > > > > > > leaves > > > > > > > > > the buffer of the transmitting interface - so that > > >retransmission > > > > can > > > > > >be > > > > > > > > > accomplished at the data link layer rather than left to >upper > > > > layers. > > > > > > > > > > > > > > > > > > Several months ago, Priscilla and I debated the 7 switch >rule. > > >If > > > > > you > > > > > > > > wanted > > > > > > > > > to search the archives for the entire thread, it was titled > > > > "What's > > > > > >the > > > > > > > > > diameter of your switched network? [7:17489]" and was >discussed > > >at > > > > > the > > > > > > > end > > > > > > > > of > > > > > > > > > August, 2001. Here's an excerpt from one of my posts >regarding > > > > the 7 > > > > > > > hop > > > > > > > > > limit: > > > > > > > > > > > > > > > > > > From other statements I've read (Cisco published material) >and > > > > from > > > > > >the > > > > > > > > > original excerpt I published, I'd imagine that the placement >of > > > > the > > > > > >root > > > > > > > > does > > > > > > > > > matter. > > > > > > > > > > > > > > > > > > "Part of this restriction is coming from the age field BPDU > > >carry: > > > > > > > > > when a BPDU is propagated from the root bridge towards the > > >leaves > > > > of > > > > > >the > > > > > > > > > tree, the age field is incremented each time it goes though >a > > > > bridge. > > > > > > > > > Eventually, when the age field of a BPDU goes beyond max >age, > > it > > > > is > > > > > > > > > discarded. Typically, this will occur if the root is too far > > >away > > > > > from > > > > > > > > some > > > > > > > > > bridges of the network. This issue will impact convergence >of > > >the > > > > > > > spanning > > > > > > > > > tree." > > > > > > > > > > > > > > > > > > I'd think that if a bridge were to be the third bridge away > > from > > > > the > > > > > > > root, > > > > > > > > > and > > > > > > > > > another switch was the third bridge on the far side of the > > root, > > >I > > > > > > > > wouldn't > > > > > > > > > expect to see any problems with MaxAge because I can't see >the > > > > root > > > > > > > being > > > > > > > > too > > > > > > > > > far from some of the bridges in the network. Now if a >bridge > > >were > > > > to > > > > > >be > > > > > > > > the > > > > > > > > > seventh, I could see how that would impose a greater delay >and > > > > > >possibly > > > > > > > > > negatively impact the MaxAge parameter. Now my question >would > > > > be... > > > > > > > does > > > > > > > > > this > > > > > > > > > really apply in today's networks or is this more of a > > limitation > > > > of > > > > > > > > > yesteryear's "software-based bridges"? > > > > > > > > > > > > > > > > > > And essentially, that's the conclusion Priscilla and I came >to > > - > > > > that > > > > > > > the > > > > > > > > 7 > > > > > > > > > hop radius doesn't really seem to apply to today's switched > > > > > > > > environments... > > > > > > > > > You might want to check with her again though - Priscilla >just > > > > > >authored > > > > > > > a > > > > > > > > new > > > > > > > > > book on troubleshooting campus networks and may updated her > > > > thinking. > > > > > > > > > > > > > > > > > > > > > > > > > > > -- Leigh Anne Chisholm (CCNP, CCDP) -----Original > > >Message----- > > > > > > > > > > From: [EMAIL PROTECTED] >[mailto:[EMAIL PROTECTED]]On > > > > Behalf > > > > > >Of > > > > > > > > > > Steven A. Ridder > > > > > > > > > > Sent: Saturday, May 18, 2002 5:16 AM > > > > > > > > > > To: [EMAIL PROTECTED] > > > > > > > > > > Subject: Re: STP and 7 hops [7:44408] > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I believe the 5-4-3 rule is for repeaters, not switches. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > ""Brian Hill"" wrote in message > > > > > > > > > > [EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > > > > > > > > > > > Steven, > > > > > > > > > > > > > > > > > > > > > > The 7 hop limit is from the root bridge, as STP >calculates > > >the > > > > > >tree > > > > > > > > from > > > > > > > > > > the > > > > > > > > > > > root. Historically, I am not sure why it's 7, but >Ethernet > > >has > > > > a > > > > > > > base > > > > > > > > hop > > > > > > > > > > > "limit" of 4 switches (5-4-3 rule), so it doesn't really > > > > matter > > > > > so > > > > > > > > much. > > > > > > > > > > The > > > > > > > > > > > reason for the 4 hop limit in Ethernet is simple: For 10 >Mb > > >or > > > > > >full > > > > > > > > > duplex > > > > > > > > > > > 100 Mb connections, the limit is mostly to reduce noise > > from > > > > the > > > > > > > > > > > amplification of the signal as it passes through the > > > > > >switches/hubs, > > > > > > > > where > > > > > > > > > > as > > > > > > > > > > > in 100Mb half-duplex connections, it is mostly to keep >the > > > > > > > propogation > > > > > > > > > > delay > > > > > > > > > > > within specs. > > > > > > > > > > > > > > > > > > > > > > Hope this helps, > > > > > > > > > > > > > > > > > > > > > > Brian Hill > > > > > > > > > > > CCNP, CCDP, MCSE 2000 (Charter Member),MCSE+I (NT4.0), > > > > > > > > > > > MCSA (Charter Member), MCP+I, MCP(21), Inet+, Net+, A+ > > > > > > > > > > > Lead Technology Architect, TechTrain > > > > > > > > > > > Author: Cisco, The Complete Reference > > > > > > > > > > > http://www.alfageek.com > > > > > ________________________ > > > > > > > > > > Priscilla Oppenheimer > > > > > http://www.priscilla.com > > ________________________ > > > > Priscilla Oppenheimer > > http://www.priscilla.com ________________________ Priscilla Oppenheimer http://www.priscilla.com Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=44769&t=44408 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
