Internet Engineering Task Force M. Pierce Internet Draft Artel draft-pierce-ieprep-pref-treat-examples-00.txt Don Choi October 2002 DISA Expires April 2003 Examples for Provision of Preferential Treatment in Voice over IP draft-pierce-ieprep-pref-treat-examples-00.txt Status of this memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress". The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt To view the list Internet-Draft Shadow Directories, see http://www.ietf.org/shadow.html. Copyright Copyright (C) Internet Society 2002. All rights reserved. Reproduction or translation of the complete document, but not of extracts, including this notice, is freely permitted. Abstract Assured Service refers to the set of capabilities used to ensure that mission critical communications are setup and remain connected. [Pierce1] describes the requirements, one of which is to provide preferential treatment to higher priority calls. IEPS refers to a set of capabilities used to provide a higher probability of call completion to emergency calls made by authorized personnel, usually from ordinary telephones. This also requires some form of preferential treatment. This memo describes some of the methods which may be applied to provide that preferential treatment. This is intended as an informational memo. Mike Pierce Expires April 2003 [Page 1] Internet Draft Examples of Preferential Treatment October 2002 Table of Contents 0. Changes...........................................................2 1. Introduction......................................................2 2. Background........................................................3 3. Potential Preferential Treatments.................................3 3.1. Reservation of Network Resources...............................3 3.1.1. RSVP.......................................................4 3.1.2. MPLS.......................................................4 3.2. Use of Higher Call Acceptance Limits...........................5 3.3. Preferential Queuing of Signaling Messages.....................6 3.4. Preferential Queuing of User Data Packets......................6 3.5. Discarding of Packets..........................................6 3.5.1. Use of DiffServ............................................6 3.5.2. Treatment for Signaling Packets............................8 3.5.3. Treatment for Voice Packets................................8 3.6. Preemption of One or More Existing Calls.......................9 4. Preemption of Some of the Resources Being Used...................10 4.1. Preemption of the Reservation.................................10 4.2. Others........................................................10 5. Security Considerations..........................................10 6. IANA Considerations..............................................10 7. References.......................................................10 8. Authors' Addresses...............................................11 0. Changes This draft was originally submitted under SIPPING, but this revision is being submitted under IEPREP to focus consideration and discussion in that WG in conjunction with the related discussions for IEPS. (SIPPING) -00 Initial version based on material removed from draft- pierce-sipping-assured-service-01. (IEPREP) -00 Added references to IEPREP in Intro. Update references. add details about packet dropping procedure. 1. Introduction [Pierce1] defines the requirements for Assured Service in support of networks requiring precedence treatment for certain calls, such as the US military network. One of those requirements is Preferential treatment, which is the following: It must be possible to provide preferential treatment to higher precedence calls in relation to lower precedence calls. Examples of preferential treatments are: - reservation of network resources for precedence calls - usage of higher Call Acceptance limits for higher precedence calls Mike Pierce Expires April 2003 [Page 2] Internet Draft Examples of Preferential Treatment October 2002 - preferential queuing of signaling messages based on precedence level - preferential queuing of user data packets based on precedence level - discarding of packets of lower precedence call - preemption of one or more existing calls of lower precedence level - preemption of some of the resources being used by a call of lower precedence level - preemption of the reservation of resources being held for other traffic Several current drafts describe the requirements for provision of the International Emergency Preparedness Scheme (IEPS). This service requires some types of preferential treatment for these calls, which can be viewed as a subset of the requirements for Assured Service listed above. These requirements include: - higher probability of call completion - lower probability of premature disconnect - distinguish IEPS data packets from other types of VoIP Packets in order to give them "priority". - Alternate path routing This informational memo describes some ways in which the above listed preferential treatments may be provided by utilizing current or new capabilities. 2. Background The requirement for Precedence Level Marking of a call setup attempt will be met by the use of the Resource Priority header defined in [Polk2]. The value carried in this header represents the relative precedence level of the call, and is used to control any of the following described procedures for providing Preferential Treatment. 3. Potential Preferential Treatments The requirement to provide preferential treatment to calls may be met by applying any combination of the following procedures: 3.1. Reservation of Network Resources Mike Pierce Expires April 2003 [Page 3] Internet Draft Examples of Preferential Treatment October 2002 This procedure involves pre-reserving certain network resources during periods when no higher precedence traffic is present so as to be prepared to handle a given level of high precedence traffic in the case of an emergency. While this method is already used in the circuit switched environment, it is less than desirable since it requires a tradeoff between the amount of wasted resources during non-emergency periods and the amount of emergency traffic which can be handled using those reserved facilities. IETF defined QoS mechanisms for packet-mode operation offer some improvement to this situation by allowing the amount of reserved resources to be adjusted. 3.1.1. RSVP RSVP may be used to establish multiple trunk groups between switching points, with each trunk group serving a different precedence level of calls. Each trunk group would be sized based on the number of simultaneous calls of that precedence level to be supported. (In this context, a trunk group refers to a facility which can support a certain number of voice connections at a certain Quality of Service level. As noted later, the number of connections can be increased with a corresponding decease in the QoS level.) With TE, the reserved sizes of these trunk groups could be adjusted during times of emergency. No preemption of these trunk groups is needed. However, reducing the size of a group to near zero would prevent further calls from using it while allowing existing calls to continue. 3.1.2. MPLS MPLS may be used to establish the equivalent of dedicated trunk groups between switching entities, enterprise network, etc. Each of these "trunk groups" could exist to support a specific precedence level of traffic between two points and could be setup using the procedures defined in [RFC3212] or those in [RFC3209]. These support the signaling of the required five levels of precedence. 3.1.2.1. Constraint-based LSP Setup using LDP [RFC3212] defines an extension to LDP to provide a constraint-based routing using MPLS. One of the constraints is based on the notion of a "priority" level for the new setup. It includes the signaling of a setup priority and a holding priority with the value of each being 0-7 (0 is the highest priority). When setting up an LSP as a trunk group to carry the traffic of one of the expected precedence levels defined in [Pierce1], the following mapping would be used: Mike Pierce Expires April 2003 [Page 4] Internet Draft Examples of Preferential Treatment October 2002 + ------------------------------------------+ | Assured Service | RFC3212 Preemption TLV | | Precedence |------------------------| | Level | SetPrio | HoldPrio | |------------------+-----------+------------| | Routine | 4 | 0 | | Priority | 3 | 0 | | Immediate | 2 | 0 | | Flash | 1 | 0 | | Flash Override | 0 | 0 | + ------------------------------------------+ This mapping prevents any preemption of a trunk group for the establishment of another. Rather, it is expected that trunk groups for all precedence levels would be initially created and remain. Only their allocated size might be changed. If actual preemption were desired, the appropriate HoldPrio values would be used. 3.1.2.2. RSVP-TE: Extensions to RSVP for LSP Tunnels As an alternative to LDP, [RFC3209] defines the use of RSVP with extensions to perform the label distribution for MPLS. It also includes the same setup and holding priorities as defined in [RFC3212]. When using RSVP as the label distribution protocol, the same mapping shown above for LDP would be used. 3.2. Use of Higher Call Acceptance Limits One aspect of preferential treatment may be provided, without resorting to preemption of calls, by allowing higher precedence calls to be setup even when they result in exceeding the engineered traffic limit on a facility (on an MPLS LSR, for example). For example, the limits for call acceptance for new calls could be set as depicted in the following table, where the engineered capacity of a route or facility is "x". A new call of each precedence level would be allowed only if the current load is within the limit shown: +------------------------------+ | Precedence Level | Capacity | | | limit of | |------------------+-----------| | Routine | .9x | | Priority | .95x | | Immediate | x | | Flash | 1.2x | | Flash-override | 1.3x | +------------------------------+ Explanation of table: In this example, a new Flash call is allowed to be setup if the current traffic load for all traffic on the Mike Pierce Expires April 2003 [Page 5] Internet Draft Examples of Preferential Treatment October 2002 facility is less than 1.2x. In the example shown in this table, Routine traffic is always prevented from using the last 10% of the capacity. The choice of the multipliers would be based on an analysis of the tradeoff between getting the high precedence level call through vs. sacrificing of QoS. It would depend on the voice encoding algorithms typically used and the end user expectations. This procedure is based on a requirement that Flash override calls should "never" be blocked. (In a probability-based system, there is no such thing as "never".) In the circuit-switched environment this could only be guaranteed by having as many circuits as there might be Flash override calls. For IP-based service, there is no fixed number of "circuits" on any facility. The "x" referred to above is only an engineering limit based on a guarantee for the provision of a certain QoS for normal traffic, i.e., Routine and Priority. This "x" may be thought of as the number of "circuits" for normal traffic. It is preferable to allow the setup of additional higher precedence calls with reduced QoS rather than blocking their setup. For example, while a particular facility may support 100 normal calls (Routine and Priority) at the guaranteed QoS, it might support 130 flash-override calls at a reduced, yet acceptable, QoS (due to packet loss) when in an emergency situation. Since the packet preferential treatment using Diff-Serv described in Section 3.4 and 3.5 could result in the discard or loss of the packets for the lower precedence calls, the higher precedence calls could still be provided a sufficient QoS even though they may have caused the engineered capacity of the route to be exceeded. The lower precedence calls will then experience higher packet discard rates or queuing delay times. If the discard rate or delay for these lower precedence calls is excessive, the end user will experience poor QoS and will likely disconnect, thereby freeing up the resources. This "encouraged disconnect" may be thought of as a "graceful preemption". 3.3. Preferential Queuing of Signaling Messages There is no plan to apply preferential queuing of signaling messages, just as this was not done in the circuit switched network. No advantage can be shown for such a procedure and it would only aggravate the problem of out-of-order messages. 3.4. Preferential Queuing of User Data Packets It is not expected that priority queuing of user data packets would provide a useful capability. It would only tend to prevent the lower priority packets from achieving the behavior required. 3.5. Discarding of Packets 3.5.1. Use of DiffServ Mike Pierce Expires April 2003 [Page 6] Internet Draft Examples of Preferential Treatment October 2002 Within DiffServ, Assured Forwarding [RFC2597] provides four classes and three drop precedences for each class (12 DSCP code points). One of these classes would be used for the signaling messages for session establishment and release. AF is not considered as being appropriate for interactive voice. Expedited Forwarding [RFC3246] defines a single class (DSCP code point) and operation, but does not include multiple drop precedences as AF does. The intention of EF is to "provide low loss, latency and jitter" and is understood to be intended for traffic such as speech, although RFC 3246 does not explicitly mention speech or voice. However, speech is less susceptible to loss than the signaling traffic and, under some traffic situations, will constitute a much larger portion of the overall load. Therefore, multiple drop precedences to alleviate overload are more appropriate to EF than they are to AF. The result of this use of DiffServ classes is that voice packets are always given priority over the signaling packets and all voice packets are treated the same. While this is the desired behavior in many cases, it is not desired in those cases in which a limited sized facility could become completely occupied by voice traffic (using EF). In this situation, further signaling messages (using AF), including those to setup new high precedence calls and those to release low precedence calls, would be lost or excessively delayed. Therefore, it is necessary to reserve a small capacity for use by the AF class which serves the signaling traffic as described in Section 2.10 of RFC 3246. For that portion of the capacity using EF for voice, the required preferential treatment for the five call precedence levels may be provided by the use of multiple drop precedence (probability) levels for packets. The procedures for the interworking of these drop precedence levels would be the same as defined currently for AF [RFC2597]. Five such levels for packet marking, using DSCPs, are necessary to provide the required functionality. In the absence of "standardized" DSCP values, local values will be assigned. Based on the definitions for AF, these levels are referred to here as: - Very low (i.e., lowest probability of being dropped) - Low - Medium - High - Very high (i.e., highest probability of being dropped) The following possible mappings are shown to illustrate the concept of using DiffServ codepoints to assist in the provision of preferential treatment to the individual packets which make up the information transfer (both the connection setup signaling and the voice transfer) of an Assured Service call. Mike Pierce Expires April 2003 [Page 7] Internet Draft Examples of Preferential Treatment October 2002 3.5.2. Treatment for Signaling Packets Consideration could also be given to utilization of different drop precedences for the signaling messages of different precedence sessions. However, using SS#7 in the PSTN as a basis, it might also be meaningful to provide different drop precedences based on type of message rather than only based on the precedence of the call. For example, for routine traffic, those messages which cause the release of sessions could be given a lower drop precedence than those which set up new sessions in order to allow such releases to take place properly under overload conditions. High precedence calls, on the other hand could use a lower drop precedence level for session setup messages than those of routine precedence calls. The following table shows what is defined for SS#7 [T1.111], including High Probability of Completion [T1.631] and MLPP [T1.619], and what might be used for SIP. (Note: The highest SS#7 Congestion Priority Level, i.e., "3", is the last to be dropped during congestion.) (Refer to draft-ietf-sipping-isup-06 for mapping of ISUP to SIP messages.) +---------------------------------+-----------------------------+ | SS#7 | SIP | +--------------------+------------+----------------+------------+ | Message | Congestion | Message | Drop | | | Priority | | Precedence | | | Level | | Level | +--------------------+------------+----------------+------------+ | Network management | 3 | ? | low | | ANM | 2 | 200 OK (INVITE)| (note) | | RLC | 2 | 200 OK (BYE) | (note) | | IAM (MLPP) | 1 or 2 | INVITE (AS) | low/medium | | IAM (HPC) | 1 | INVITE (IEPS) | low/medium | | ACM | 1 | 18x | medium | | CPG | 1 | 100 Trying/18x | medium | | REL | 1 | BYE | low | | IAM (normal) | 0 | INVITE (normal)| high | | Others | 0 | | | +--------------------+------------+----------------+------------+ Note: Within SIP, all ACKs would need to have the same preferential treatment as the message they are acknowledging. 3.5.3. Treatment for Voice Packets This example is for the case of the use of DiffServ to provide the packet forwarding preferential treatment through multiple drop precedence levels. Each packet containing user data (voice) is marked with a unique DiffServ codepoint to indicate one of the following levels and resulting treatment: Mike Pierce Expires April 2003 [Page 8] Internet Draft Examples of Preferential Treatment October 2002 +-----------------------------------+-----------------+ | Precedence | Indication in user | Drop if current | | Level | voice packets | queue is more | | |--------------------| than -- % full | | | Class | Drop | | | | | precedence | | |--------------+-------+------------|-----------------+ |Routine | EF | Very high | 80% | |Priority | EF | High | 90% | |Immediate | EF | Medium | 100% | |Flash | EF | Low | 110% | |Flash Override| EF | Very low | 120% | +-----------------------------------+-----------------+ All voice traffic is then served by a single instance of EF, and served by a single (strict FIFO) queue. This results is an equal treatment in terms of delay variation (often called "jitter") for all precedence levels for those packets which are delivered, but achieves this by selective packet discard. The discard may use a simple tail dropping algorithm as shown in the above table or a form of "Random Early Detection" as described in [RFC2309] to drop some packets before the queue actually reaches the fill shown above. However, since the packets in this queue are not using TCP and can not be bursty or "agressive", there appears to be no advantage gained by the complexity of early detection and random dropping. Note that "queue full" here refers to the engineered limit, that is, the limit which should be applied in order to meet the requirements of the EF PHB and the desired QoS. Since this calculation is based on probabilities of achieving a certain target QoS, it can be temporarily exceeded as described in the following section. 3.6. Preemption of One or More Existing Calls The procedures described above for use of higher call acceptance limits and selective discard of voice packets based on the precedence level of the call reduce or eliminate the need to perform preemption of existing calls within the IP domain. The statistical nature of packet transmission makes it possible to "squeeze" an additional high precedence call into an already "full" facility, as illustrated in the previous section. It should be noted that, in the extreme case, these procedures would result in the same effect as preemption, since the resources of the lower precedence call would be so severely degraded (via packet loss) that communication would be impossible and the users would likely disconnect. When interworking with circuit switched portions of the telecommunications network, preemption procedures are still required within transport facilities which are based on fixed numbers of circuits. In some cases, this preemption must result in specific procedures being applied in the packet portion, such as notifications of preemption. Mike Pierce Expires April 2003 [Page 9] Internet Draft Examples of Preferential Treatment October 2002 4. Preemption of Some of the Resources Being Used The "preemption" of some of the resources being used by lower precedence traffic may be accomplished through the packet discard described above. 4.1. Preemption of the Reservation Based on traffic engineering, the amount of resources allocated to reserved paths (e.g., MPLS or RSVP) could be adjusted. For example, when an emergency situation occurs, the need for more resources to support higher priority traffic could be recognized. The existing LSPs could be changed using the procedures of [RFC3214] to allow the size of those LSPs supporting the higher priority traffic to be increased while others are decreased. 4.2. Others There may be other procedures which could be used to provide the required preferential treatments. 5. Security Considerations The security considerations are covered in [Pierce1]. 6. IANA Considerations It is not expected that there will be any IANA involvement in support of provision of Preferential Treatment for Assured Service beyond what is described in [Polk2]. 7. References [RFC2205] "Resource ReSerVation Protocol (RSVP)", September 1997 [RFC2309] "Recommendations on Queue Management and Congestion Avoidance". April 1998 [RFC2597] "Assured Forwarding PHB Group", June 1999. [RFC3246] "An Expedited Forwarding PHB", March 2002. [RFC2751] "Signaled Preemption Priority Policy Element", January 2000. [RFC3209] "RSVP-TE: Extensions to RSVP for LSP Tunnels", December 2001. [RFC3212] "CR-LDP: Constraint-based LSP Setup using LDP", January 2002. Mike Pierce Expires April 2003 [Page 10] Internet Draft Examples of Preferential Treatment October 2002 [RFC3214] "LSP Modification Using CR-LDP", January 2002. [RFC3261] "SIP: Session Initiation Protocol", June 2002. [T1.111] ANSI T1.111-2001, "Signalling System No. 7 (SS7) - Message Transfer Part". [T1.619] ANSI T1.619-1992 (R1999) "ISDN - Multi-Level Precedence and Preemption (MLPP) Service Capability". [T1.631] ANSI T1.631-1993 (R1999) "Telecommunications - Signalling System No. 7 (SS7) - High Probability of Completion (HPC) Network Capability". [Pierce1] draft-pierce-ieprep-assured-service-req-00, "Requirements for Assured Service Capabilities in Voice over IP", October 2002 [Polk2] draft-polk-sipping-resource-00, "SIP Communications Resource Priority Header", February 2002. [Talauliker] draft-talauliker-ieprep-diffserv-00, "Internet Emergency Preparedness Services in a Differentiated Services Domain", June 2002. 8. Authors' Addresses Michael Pierce Artel 1893 Preston White Drive Reston, VA 20191 Phone: +1 410.817.4795 Email: pierce1m@ncr.disa.mil Don Choi DISA 5600 Columbia Pike Falls Church, VA 22041-2717 Phone: +1 703.681.2312 Email: choid@ncr.disa.mil Full Copyright Statement Copyright (c) The Internet Society (2002). All Rights Reserved. This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Mike Pierce Expires April 2003 [Page 11] Internet Draft Examples of Preferential Treatment October 2002 Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English. The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assigns. This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Mike Pierce Expires April 2003 [Page 12]