The MPLS WG Archive[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index][Thread Index][Author Index][Subject Index] Questions related to Class-Type (in DiffServ-TE context)
Hi! Thanks for your response. But, I am still not clear about the purpose of sending the Class-Types in the (RSVP-TE & CR-LDP) signaling messages. In your reply you pointed out that PHB<-->Class-Type mapping has not yet been defined. Is it really author's intention to define, yet another higher layer grouping of PHBs (i.e a Class-Type consists of several PHBs ? If the administrator really wanted to define a class for Real-time, or Low-Loss, can't he define a new PHB ?) I thought the draft diff-te-reqts-00, implies that the main purpose of the introduction of Class-Type was to reduce the number of IGP messages required to communicate the per-diffserv class bandwidth information user configured on a LSR. Therefore, I understand that there need for changes to the IGPs, but I am unclear about the purpose of the Class-Type in the signaling messages. Your help in clarifying this will be appreciated. Thanks, sanjay -----Original Message----- From: Vadim Suraev [mailto:Vadim_Suraev@KereniX.com] Sent: Wednesday, February 21, 2001 5:01 PM To: 'Choudhury, Sanjaya' Cc: 'mpls@UU.NET' Subject: RE: Questions related to Class-Type (in DiffServ-TE context) Sanjay, if I'm not mistaken: > Questions regarding Definition of Class Types: > ------------------------------------------------------- > Q1. Who defines the Class-Types (0,1,2,3) ? If you ask about which PHBs are associated with which Class Type - I think, this is task of network operator to define, which PHB is in which Class Type. I think, the signaling for PHB(s)<-->Class Types mapping is not defined, but there is - ability to signal which PHBs LSP carries (draft-ietf-mpls-diff-ext-08.txt); - ability to allocate BW for particular LSP from unreserved BW of certain Class Type (draft-ietf-mpls-diff-te-ext-01.txt). So, I think, operator have to define this in such way, that particular set of PHBs will share BW from BW pool (unreserved BW), associated with same Class Type and particular PHB will not allocate BW from more than one Class Type. > Q2. If these Class-Types are administrator defined and is configured > in all the LSR of the DiffServ-TE domain, why is the Class-Type > signaled ? > (Signaling messages already signals the PSC(s) being carried by > the LSP, can't one determine the class-type from local configuration?) I think, this allows you to avoid configuration of PHBs<-->Class Type mapping (and checking at LSP setup) at each LSR. > Q3. The diff-te-reqts-00 draft states that "per Class TE" can be > viewed as a special case of per Class-Type TE, where each Class-Type > is degenerated into a single Diff-Serv class. In this cases, is the > Class-Type information in the signaling messages redundant ? I think, it is not redundant because when you insert Class Type information into signaling message, you tell to each LSR on the path that you wish to allocate BW from BW pool of the Class Type. For example, you may associate EF with Class Type 1 and AFxx with Class Type 2. Then Class Type 1 includes only one DiffServ class (EF) and BW which should be allocated for LSP, which carries this class (EF), should be taken from BW pool (unreserved BW) of Class Type 1. Therefore, Class Type 1 should be signaled for LSP, which carries EF, in this case. Also, you may setup LSP for AF11 only rather than for AFxx. Since AF11 is subset of AFxx (it means here that AF11 may share BW with other AFs) and AFxx is associated logically with Class Type 2, you must to to signal the Class Type 2. > Questions regarding the Admission Control of L-LSP in DiffServ-TE > environment. > > ------------------------------------------------------------------------- > Q1. The mpls-diff-ext-07 draft, indicates that the signaled > bandwidth can be used by the LSR to perform admission control of the signaled LSP > over the DiffServ resources provisioned for the relevant PSC(s). > The diff-te-ext-01 draft, states that if the CLASSTYPE object is not > present in the Path message, the LSR must associate the Class-Type 0 > to the LSP. This implies that one is expected to do admission control > against the resource available for Class-Type 0. > How can I read these two together ? I think, mpls-diff-ext-08 means, that each LSR is provisioned initially (using SNMP, for example) with per-PSC (per Class Type) reservable BW (i.e. all link's reservable BW may be divided into per PSC (per Class Type) BW pools). So, if you do not include the CLASSTYPE object into the Path message, LSP is automatically associated with Class Type 0 and (if I'm not mistaken), LSP's signaled PSC (described in mpls-diff-ext-08) is expected to be associated (logically) with Class Type 0. Otherwise (if LSP's signaled PSC is associated logically with other Class Type), LSP will try to allocate BW from unreserved BW, intended to other PSC(s). > Pre-emption and Class-Type > > -------------------------------------------------------------------------- > Q1. Based on the diff-te-reqts-01, can I assume that the pre-emption > should NOT work across the class-types (unlike the suggestion in the > 00 version of the draft) ? Do you mean the preemption priority levels? If so, I think it remains (there is draft-ietf-ospf-diff-te-00.txt, describes unreserved BW advertisements for different preemption priority levels). |
|