RE: [LinkSec] RE: [802.1] P802.1ad Task Group ballot disposition

Thread Links	Date Links
Thread Prev	Thread Next	Thread Index	Date Prev	Date Next	Date Index

I disagree that the technical details in 802.1ad D1 require that an Island be limited to 4094 service instances. However this is a lot to expect someone to figure out from 802.1ad D1 which needs to be consistent in its description to allow this to be figured out - plainly it is not. One place it says one thing, and another in another.

I do believe they require that any Bridge in the island carry no more than 4094 service instances, which is the same thing. Arichly connected core together with multiple spanning tree instances and some regional locality of traffic (no more than I have seen in practice) should get the number to about 100,000 max. I think this is way more than anyone wants for a flat bridged net, but am prepared to be convinced otherwise.

I think desirability will be a large extent predicated on (perceived) cost. If the existing technology can be used to support a few thousand service instances then a more (per port) expensive interconnect can distribute its costs across these few thousand. Put against a different technology without the volume underpinned by enetrprise use this will be attractive to many.

I don't expect anyone to agree on these costs but I personally suspect that no one is going to find a large VID size that will be successful against such an architecture, though those who want to back completely different technologies will enjoy a VID size discussion.

-----Original Message-----
From: owner-stds-802-linksec@majordomo.ieee.org [mailto:owner-stds-802-linksec@majordomo.ieee.org]On Behalf Of Marc Holness
Sent: Friday, July 18, 2003 11:56 AM
To: mick_seaman@ieee.org; 'Dirceu Cavendish'; stds-802-1@ieee.org; stds-802-linksec@ieee.org
Subject: [LinkSec] RE: [802.1] P802.1ad Task Group ballot disposition

The number of local service instances supportable within an Island (as described in 802.1ad D1 draft) is still limited to 4094 (maximum). Yes, given a network composed of multiple small Islands and Interconnect Media, the overall network service identification space (e.g., VPN ID space) can be increased.

My read on comment #44 is a request to increase the number of local service instances within an Island. There is no suggestion of a globally unique ID.

It is probably highly desirable to potential users of this technology if they could deploy a single Island (or a minimum set) that can support a fairly large service instance space rather than multiple Islands and an Interconnect Media, etc.

-----Original Message-----
From: Mick Seaman [mailto:mick_seaman@ieee.org]
Sent: Wednesday, July 16, 2003 5:06 PM
To: 'Dirceu Cavendish'; stds-802-1@ieee.org; stds-802-linksec@ieee.org
Subject: RE: [802.1] P802.1ad Task Group ballot disposition

The connection between the number of bits that need to be in an S-VID in the front of any particular frame and the number of bits needed to identify all the points of attachment to the network that receive service is not quite as obvious as it may appear initially:

(a) any given provider bridged network that is purely bridged can have S-VIDbits**2 service instances passing through any given bridge

- for the purposes of illustration lets assume that the network is designed around a core of 16 full mesh connected colocation sites that connect a number of network regions, under the rather unlikely and testing case (because traffic tends to be more regionally concentrated than this) that desired connectivity is randomly distributed (but dominated by point to point which is realistic so far as my own market experience goes), then the network capacity provided by a PVID of 12 bits is 4094/2 * 16 = 32,752 service instances. If as is likely from experience, most service instances connect within the region (particularly to the ISPs regional drop) this number is greater.

(b) any provider bridged network that uses other technology to connect pure bridging regions (a good choice from a cost point of view, get the cost down as low as possible where the numbers are greatest then use fancier technology to connect regions of these numbers) can have 4094 * number of connections to the core technology.

(c) where one customer distributes services to a number of subscribers, hundreds if not tens of thousands of these subscribers can be accomodated on one VLAN (see Norm's reply). (There is some interaction of this point with point (b) so I mustn't double count, but it can be used to multiply the gains of type (a) without restriction.

-----Original Message-----
From: owner-stds-802-1@majordomo.ieee.org
[mailto:owner-stds-802-1@majordomo.ieee.org]On Behalf Of Dirceu Cavendish
Sent: Wednesday, July 16, 2003 10:20 AM
To: 'Tony Jeffree'; stds-802-1@ieee.org; stds-802-linksec@ieee.org
Subject: RE: [802.1] P802.1ad Task Group ballot disposition

Comment about "comment 44" and its response. The issue is the P-VID space. The comment is that more bits are needed for SERVICES envisaged by NTT's network. The response seems to imply that a too large network is not desirable. One may have a tiny little provider network in a very densely populated area (common in Japan big cities), and yet with hundred of thousands of "service instances"...

Dirceu Cavendish
NEC Labs America
10080 North Wolfe Road Suite SW3-350
Cupertino, CA 95014
Tel: 408-863-6041 Fax: 408-863-6099

-----Original Message-----
From: owner-stds-802-1@majordomo.ieee.org
[mailto:owner-stds-802-1@majordomo.ieee.org] On Behalf Of Tony Jeffree
Sent: Tuesday, July 15, 2003 5:37 AM
To: stds-802-1@ieee.org; stds-802-linksec@ieee.org
Subject: [802.1] P802.1ad Task Group ballot disposition