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

[STDS-802-16] FW: [STDS-802-16] [NETMAN] What is "BS Sector Mac Address" role



Krzysztof suggested I forward my response to the reflector...
Enjoy!

 

--

Mike Geipel
Principal Engineer
Axxcelera Broadband Wireless
1600 East Parham Road

Richmond, VA 23228
Direct: 804-864-4125



From: Mike Geipel   
Sent: Wednesday, February 16, 2005 12:49 PM
To: 'Krzysztof Dudzinski'
Subject: RE: [STDS-802-16] [NETMAN] What is "BS Sector Mac Address" role

Hi Krzysztof,
 
Yes, I saw the message from Jeff, but I wasn't aware of that connection either.
It would certainly be convenient for BSID to be manufacturer-supplied MAC address, especially in 802.16e...
 
Looking at the spec ...
... if BSID is issued by operator, it cannot, in general, be a MAC address.
 
In 802.16-2004, section 6.3.2.3.2:
"The Base Station ID is a 48-bit long field identifying the BS. The Base Station ID shall be programmable. The most significant 24 bits shall be used as the operator ID. This is a network management hook that can be combined with the Downlink Channel ID of the DCD message for handling edge-of-sector and edge-of-cell situations."
This suggests that the BSID is unique to basestation, not necessarily per-sector.
 
The only place where I saw the right semantics WRT sectors, was in descriptions of Sector ID in compressed DL-MAPs, SCa on page 403 section 8.2.1.8.1, and OFDMA on page 549, section 8.4.5.6.1.
 
We should probably define the following in the Corrigenda Glossary:
  Base Station Identifier
  Operator Identifier
  Sector Identifier
 
Although the sizes are right, there is no statement that the most significant 24 bits are an IEEE OUI, and that least significant 24 bits are guaranteed unique (for that operator).  Even if we were to fix the Corrigenda to say so, I'm not sure how to enforce this.  The consequences of not getting this right could lead to really bad forwarding behavior in certain deployments.
 
If BSID is manually issued by operators, I would likely argue that it is not the ifPhysAddress.
 
 
I'm not sure I understood the context for your other question about interface port:
 - I would think that each sector should be considered a separate port as defined in RFC1493 (BRIDGE-MIB)
 - Each downlink channel should be a separate interface, as defined in RFC2863 (IF-MIB)
I think it would be prudent for the 802.16f spec to define how interfaces are stacked, according to RFC2863 (and possibly RFC2864).
 
I hope this helps.
 

--

Mike Geipel
Principal Engineer
Axxcelera Broadband Wireless
1600 East Parham Road

Richmond, VA 23228
Direct: 804-864-4125

 


From: Krzysztof Dudzinski   
Sent: Wednesday, February 16, 2005 7:32 AM
To: Mike Geipel
Subject: RE: [STDS-802-16] [NETMAN] What is "BS Sector Mac Address" role

Mike,
 
Thanks for this.
Our plan is to provide MacAddresses as specified in the MIB document. But this triggers some more questions and leads to confusion. We may need to clarify the language used in MIB document to be more specific. Here is an example of two. Do you have an opinion about that?
 
- What is the relationship between BSID and BS (Sector) Mac Address? Jeff Mandin comented about it but I don't see how we can use BSID as a PhysAddress in ifTable. Mac Address is guaranteed unique and BSID not yet. Mac Address is allocated by manufacturer, BSID by user (operator).
- Where is the interface port we assign PhysAddress to? It is easy with Ethernet but our .16 interface port is rather virtual, isn't it?
 
Regards,
Krzysztof
 


From: Mike Geipel
Sent: 13 February 2005 22:28
To: Krzysztof Dudzinski
Subject: RE: [STDS-802-16] [NETMAN] What is "BS Sector Mac Address" role

All SNMP implementations are required to conform to RFC1213.
 
The RFC1213-MIB defines a textual convention for PhysAddress ::= OCTET STRING.
It also defines ifPhysAddress OBJECT-TYPE SYNTAX PhysAddress ::= { ifEntry 6 }.
 
The latest version of the IF-MIB [RFC2863] does not change this, so this field should be there (whether it is populated or not).
 
Conceptually, each MAC SAP should (by definition) have a MAC address.
So, if there is no SAP needed for a given interface, I suppose it can still bridge frames without a unique MAC address.
There are, however, pragmatic reasons to specify a MAC address for each ifEntry.
 
 
In answer to question (1):
Without a MAC address defined for each interface, Network Management software will be unable to auto-discover the layer-2 topology accurately using SNMP.  (NNM, for example, uses RFCs 1493, 1515, 2108 for this purpose.)  Managed devices that populate their bridge tables, can then be interrogated by management software for each interface.
 
 
In 802.1D-2004, Section 6 states that "Mac Bridges interconnect the separate IEEE 802 LANs that compose a Bridged Local Area Network by relaying and filtering frames between separate MACs of the bridged LANs."  By this definition, I'd argue that the BS is a "Mac Bridge" with at least one network interface (802.3/Ethernet?) and at least one RF interface (802.16 sector).
(The Corrigenda now seems to mandate BS support for both Ethernet CS and IP-over-Ethernet CS.)
 
In 802.1D-2004, Section 7.12.2 states that:
"The individual MAC Entity associated with each Bridge Port shall have a separate individual MAC Address."  And this is reiterated in Section A., item ADDR-1.
 
 
BTW, why not?  MAC addresses are cheap.  Once you've paid for a mechanism to serialize for a single MAC address (like the BS Ethernet interface that must have a SAP), it is easy to derive (say) ten more MAC addresses in a block.
 

--

Mike Geipel
Principal Engineer
Axxcelera Broadband Wireless
1600 East Parham Road

Richmond, VA 23228
Direct: 804-864-4125

 

 


From: Krzysztof Dudzinski [mailto:KDudzins@AIRSPAN.COM]
Sent: Friday, February 11, 2005 1:15 PM
To: STDS-802-16@LISTSERV.IEEE.ORG
Subject: [STDS-802-16] [NETMAN] What is "BS Sector Mac Address" role

Hi All,

 

As described “802.16_2004_TGF.pdf, draft2, section 9.3.2.2” we decided to manage wireless man “BS Sectors” as network interface. As such there is one row defined in the interface table (ifTable) per “BS Sector” along with required attributes e.g. ifType (propBWAp2Mp), IfPhysAddress (Mac Address of BS Sector) etc.

Two questions arise:

 

1. Can anybody think of the scenario where the Mac Address of “BS Sector” would be actually used?

2. Can the ifPhysAddress be filled with zero length octet string as suggest the description of this object in the IF-MIB (see extract below) if standard doesn’t define the use of it?

 

Unlike SS Mac Address the Mac Address of BS (Sector) is not defined in the standard 802.16-2004 at all. The BS Mac Address is mentioned in the 802.16e/D5a in the context of Neighbor advertisement message. But in this context it is not sure what Mac Address the document refers to (what type of interface in the interface table).

Nevertheless the issue is with 802.16f MIB, which is designed for fixed operation.

 

Extract form IF-MIB:

ifPhysAddress OBJECT-TYPE

    SYNTAX      PhysAddress

    MAX-ACCESS  read-only

    STATUS      current

    DESCRIPTION

            "The interface's address at its protocol sub-layer.  For

            example, for an 802.x interface, this object normally

            contains a MAC address.  The interface's media-specific MIB

            must define the bit and byte ordering and the format of the

            value of this object.  For interfaces which do not have such

            an address (e.g., a serial line), this object should contain

            an octet string of zero length."

    ::= { ifEntry 6 }

Regards,

Krzysztof Dudzinski

Airspan Networks