| Audio/Video Transport Working Group | G. Hunt |
| Internet-Draft | P.J. Arden |
| Intended status: Informational | BT |
| Expires: September 15, 2011 | Q. Wu, Ed. |
| Huawei | |
| March 14, 2011 |
Monitoring Architectures for RTP
draft-hunt-avtcore-monarch-01.txt
This memo proposes an architecture for extending RTCP with a new RTCP XR (RFC3611) block type to report new metrics regarding media transmission or reception quality, as proposed in RFC5968. This memo suggests that a new block should contain a single metric or a small number of metrics relevant to a single parameter of interest or concern, rather than containing a number of metrics which attempt to provide full coverage of all those parameters of concern to a specific application. Applications may then "mix and match" to create a set of blocks which covers their set of concerns. Where possible, a specific block should be designed to be re-usable across more than one application, for example, for all of voice, streaming audio and video.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at http://datatracker.ietf.org/drafts/current/.
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."
This Internet-Draft will expire on September 15, 2011.
Copyright (c) 2011 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
Service providers and network providers today suffer from lack of good service that can monitor the performance at the user's home, handset or remote office. Without service performance metrics, it is difficult for network operators to properly locate the problem and solve service issues before problems impact subscriber/end user. The resolution generally involves deploying costly field network technician to conduct on-site troubleshooting and diagnostics. By reducing the expensive deployments with more automated remote monitoring capabilities, network operators can save significant costs, reduce mean time to repair and provider a better service offering.
As more users and subscribers rely on real time application services, uncertainties in the performance and availability of these services are driving the need to support new standard methods for gathering performance metrics from RTP applications. These rapidly emerging standards, such as RTCP XR [RFC3611]and other RTCP extension to Sender Reports(SR), Receiver Reports (RR) [RFC3550]are being developed for the purpose of collecting and reporting performance metrics from endpoint devices that can be used to correlate the metrics, provide end to end service visibility and measure and monitor QoE.
However the proliferation of RTP/RTCP specific metrics for transport and application quality monitoring has been identified as a potential problem for RTP/RTCP interoperability, which attempt to provide full coverage of all those parameters of concern to a specific application. Since different applications layered on RTP may have some monitoring requirements in common, therefore these metrics should be satisfied by a common design.
The objective of this document is to define an extensible RTP monitoring framework to provide a small number of re-usable QoS/QoE metrics which facilitate reduced implementation costs and help maximize inter-operability. [RFC5968] has stated that, where RTCP is to be extended with a new metric, the preferred mechanism is by the addition of a new RTCP XR [RFC3611] block. This memo assumes that any requirement for a new metric to be transported in RTCP will use a new RTCP XR block.
This memo is informative and as such contains no normative requirements.
The RTP monitoring architecture comprises the following three functional components shown below:
RAQMON Report Wrapper (RRW) is a functional component that acts as a source of information gathered for monitoring purposes. It also can be referred to as "Monitoring Client". The end system that source RTP streams, or an intermediate-system that forwards RTP packets to End-devices can be envisioned to act as RRWs within the RTP monitoring architecture.
A RAQMON Report Collector (RRC) is a functional component that act as monitoring server or monitoring center. It collects statistics from multiple RRWs, analyzes them, stores such information reported by RTCP XR or other RTCP extension appropriately as base metric or calculates composite metric. RRC is envisioned to be a middleware like RTP translator, Multipoint Conferencing Bridge, Distribution Source, serving an administrative domain defined by the network administrator.
The RAQMON Metric Block exposes real time Application Quality information in the report block format to RRC and Network Management Applications. The RTCP or RTCP XR can be extended to convey such information to accommodate the RTP monitoring architecture.
+-------------------+
| RTP Sender |
| +--------------+ |
| |RAQMON Report |--- --------|
| |Wrapper | | |
| +--------------+ | |
|+-----------------+| | +-----------+
||Application || | |Management |
||-streaming video || | |Application|
|---|-VOIP || | +------\----+
| ||-video conference|| 5 | |
| ||-telepresence || | |6
| ||-ad insertion || | +-------|-----+
| |+-----------------+| ------->|RAQMON Report|
| +-------------------+ ------->| Collector |
| Report Block | +----------\--+
| transported over | Report Block |
| RTCP extension | transported over|5
| 1 | RTCP XR |
| +------ -----------------+ | +--------------|---- ----+
| | RTP System | | | RTP Receiver >--4-|--- |
| | +--------------+ | 5 | | +--------------+ | |
| | |RAQMON Report --------------| | |RAQMON Report |<-- |
| | | Wrapper | | | | Wrapper |<------|
| | +--------------+ | | +-------\------+ ||
| | | | | ||
| | | | |2 ||
| | +-----------------+ | | +-------/---------+ ||
| | |Application | | | |Application | ||
| | |-streaming video | | | |-streaming video | ||
| | |-VOIP | | 1 | |-VOIP | 3|
---->-Video conference|--------------->|-Video conference ||
| |-Telepresence | | | |-Telepresence | ||
| |-Ad insertion | | | |-Ad insertion | ||
| +-----------------+ | | +-----------------+ ||
| +-----------------+ | | +-----------------+ ||
| |Transport | | | |Transport | ||
| |-IP/UDP/RTP | | | |-IP/UDP/RTP >---||
| |-IP/TCP/RTP | | | | -IP/TCP/RTP | |
| |-IP/TCP/RTSP/RTP | | | |-IP/TCP/RTSP/RTP | |
| +-----------------+ | | +-----------------+ |
+------------------------+ +------------------------+
The basic RTCP Reception Report (RR) conveys reception statistics in metric block report format for multiple RTP media streams including [RFC3611] supplement the existing RTCP packets and provide more detailed feedback on reception quality in several categories:
The RTCP XRs
There are also various other scenarios in which it is desirable to send RTCP Metric reports more frequently. The Audio/Video Profile with Feedback [RFC4585]extends the standard A/V Profile[RFC3551] to allow RTCP reports to be sent early provided RTCP bandwidth allocation is respected. There are four use cases but are not limited to:
Measured data at the application level, i.e., QoE related parameters which focus on quality of content rather than network parameters. These include but are not limited to:
Measured data at the transport level. These include but are not limited to:
Measured data from application residing in that device. These include but are not limited to:
The aim of "monitoring every packet" is to ensure that the information reported is not dependent on the application. In this scheme, RTP systems will report arrival data for each individual RTP packet. RTP (or other) systems receiving this "raw" data may use it to calculate any preferred heuristic metrics, but such calculations and the reporting of the results (e.g. to a session control layer or a management layer) are outside the scope of RTP and RTCP.
There are several potentially useful metrics which rely on the accumulation of a histogram in real time, so that a packet arrival results in a counter being incremented rather than in the creation of a new data item. These metrics may be gathered with a low and predictable storage requirement. Each counter corresponds to a single class interval or "bin" of the histogram. Examples of metrics which may be accumulated in this way include the observed distribution of packet delay variation, and the number of packets lost per unit time interval.
Different networks may have very different expected and achieved levels of performance, but it may be useful to fix the number of class intervals in the reported histogram to give a predictable volume of data. This can be achieved by starting with small class intervals ("bin widths") and automatically increasing the width (e.g. by factors of two) if outliers are seen beyond the current upper limit of the histogram. Data already accumulated may be assigned unambiguously to the new set of bins, given some simple conditions on the relationship between the old and new origins and bin widths.
A significant disadvantage of the histogram method is the loss of any information about time-domain correlations between the samples which build the histogram. For example, a histogram of packet delay variation provides no indication of whether successive samples of packet delay variation were uncorrelated, or alternatively that the packet delay variation showed a highly-correlated low-frequency wander.
An entity which both monitors the packet stream, and has sufficient knowledge of the application to know when transport impairments may have degraded the application's performance, may choose to send exception reports containing details of the transport impairments to a receiving system. The crossing of a transport impairment threshold, or some application-layer event, would trigger such reports. RTP end systems and mixers are likely to contain application implementations which may, in principle, identify this type of exception.
It is likely that RTP translators will not contain suitable implementations which could identify such exceptions.
On-path devices such as routers and switches are not likely to be aware of RTP at all. Even if they are aware of RTP, they are unlikely to be aware of the RTP-level performance required by specific applications, and hence they are unlikely to be able to identify the level of impairment at which exceptional transport conditions may start to affect application performance.
This type of monitoring typically requires the storage of recent data in a FIFO (e.g. a circular buffer) so that data relevant to the period just before and just after the exception may be reported. It is not usually helpful to report transport data only from the period following an exception event detected by an application. This imposes some storage requirement (though less than needed for Option 1). It also implies the existence of additional cross-layer primitives or APIs to trigger the transport layer to generate and send its exception report. Such a capability might be considered architecturally undesirable, in that it complicates one or more interfaces above the RTP layer.
This is a business-as-usual option which suggests that the current approach should not be changed, based on the idea that previous application-specific approaches such as that of [RFC3611] were valid. If a large category of RTP applications (such as VoIP) has a requirement for a unique set of transport metrics, arising from its different requirements of the transport, then it seems reasonable for each application category to define its preferred set of metrics to describe transport impairments. We expect that there will be few such categories, probably less than 10.
It may be easier to achieve interworking for a well-defined set of application-specific metrics than it would be in the case that applications select a profile from a palette of many independent re- usable metrics.
Issues that have come up in the past with extensions to RTCP or RTCP XR include (but are probably not limited to) the following:
Different applications using RTP for media transport certainly have differing requirements for metrics transported in RTCP to support their operation. For many applications, the basic metrics for transport impairments provided in RTCP SR and RR packets [RFC3550] (together with source identification provided in RTCP SDES packets) are sufficient. For other applications additional metrics may be required or at least sufficiently useful to justify the overheads, both of processing in endpoints and of increased session bandwidth. For example an IPTV application using Forward Error Correction (FEC) might use either a metric of post-repair loss or a metric giving detailed information about pre-repair loss bursts to optimise payload bandwidth and the strength of FEC required for changing network conditions. However there are many metrics available. It is likely that different applications or classes of applications will wish to use different metrics. Any one application is likely to require metrics for more than one parameter but if this is the case, different applications will almost certainly require different combinations of metrics. If larger blocks are defined containing multiple metrics to address the needs of each application, it becomes likely that many different such larger blocks are defined, which becomes a danger to interoperability.
To avoid this pitfall, this memo proposes the use of small RTCP XR metrics blocks each containing a very small number of individual metrics characterising only one parameter of interest to an application running over RTP. For example, at the RTP transport layer, the parameter of interest might be packet delay variation, and specifically the metric "IPDV" defined by [Y1540]. See Section 8 for architectural considerations for a metrics block, using as an example a metrics block to report packet delay variation.
Any measurement must be identified. However if metrics are delivered in small blocks there is a danger of inefficiency arising from repeating this information in a number of metrics blocks within the same RTCP packet, in cases where the same identification information applies to multiple metrics blocks.
An instance of a metric must be identified using information which is likely to include most of the following:
[Editor's note: this set of information overlaps with, but is more extensive than, that in the union of similar information in RTCP RR packets. Should we assume that RR information is always present if XR is sent, and that measurement intervals are exactly coincident? If so, state assumption and remove overlaps. What were the design considerations which led to the additional information *not* being present in RRs? The reason for the additional information here is the perceived difficulty of "locating" the *start* of the RTP session (sequence number of 1st packet, duration of interval applicable to cumulative measurements) using only RR. Is this a misconception? It leads to redundant information in this design because equivalent information is provided multiple times, once in *every* identification packet. Though this ensures immunity to packet loss, the design is ugly and the overhead is not completely trivial.]
This section proposes an approach to minimise the inefficiency of providing this identification information, assuming that an architecture based on small blocks means that a typical RTCP packet will contain more than one metrics block needing the same identification. The choice of identification information to be provided is discussed in [IDENTITY] (work in progress).
The approach is to define a stand-alone block containing only identification information, and to tag this identification block with a number which is unique within the scope of the containing RTCP XR packet. The "containing RTCP XR packet" is defined here as the RTCP XR header with PT=XR=207 defined in Section 2 of [RFC3611] and the associated payload defined by the length field of this RTCP XR header. The RTCP XR header itself includes the SSRC of the node at which all of the contained metrics were measured, hence this SSRC need not be repeated in the stand-alone identification block. A single containing RTCP XR packet may contain multiple identification blocks limited by the range of the tag field. Typically there will be one identification block per monitored source SSRC, but the use of more than one identification block for a single monitored source SSRC within a single containing RTCP XR packet is not ruled out.
There will be zero or more metrics blocks dependent on each identification block. The dependence of an instance of a metrics block on an identification block is established by the metrics block's having the same numeric value of the tag field as its identification block (in the same containing RTCP XR packet).
Figure 2 below illustrates this principle using as an example an RTCP XR packet containing four metrics blocks, reporting on streams from two sources. The measurement identity information is provided in two blocks with Block Type NMI, and tag values 0 and 1 respectively.
Note: in this example, RTCP XR block type values for four proposed new block types (work in progress) are given as NMI, NPDV, NBGL and NDEL. These represent numeric block type codepoints to be allocated by IANA at the conclusion of the work.
Each of these two identity blocks will specify the SSRC of one of the monitored streams, as well as information about the span of the measurement. There are two metrics blocks with tag=0 indicating their association with the measurement identity block which also has tag=0. These are the two blocks following the identity block with tag=0, though this positioning is not mandatory. There are also two metrics blocks with tag=1 indicating their association with the measurement identity block which also has tag=1, and these are the two blocks following the identity block with tag=1.
[Editor's note: if we mandated that metrics blocks associated with an identity block must always follow the identity block we could save the tag field and possibly simplify processing. Is this preferable to cross-referencing with a numeric tag?]
In the example, the block types of the metrics blocks associated with tag=0 are BT=NPDV (a PDV metrics block) and BT=NBGL (a burst and gap loss metrics block). The block types of the metrics blocks associated with tag=1 are BT=NPDV (a second PDV metrics block) and BT=NDEL (a delay metrics block). This illustrates that:
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|V=2|P|reserved | PT=XR=207 | length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| SSRC of RTCP XR packet sender |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| BT=NMI |0|tag=0| resv | block length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| SSRC of stream source 1 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
. ...measurement identity information, source 1... .
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| BT=NPDV |I|tag=0|pdvtyp | block length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
. ...PDV information for source 1... .
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| BT=NBGL |I|tag=0| resv | block length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
. ...burst-gap-loss information for source 1... .
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| BT=NMI |0|tag=1| resv | block length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| SSRC of stream source 2 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
. ...measurement identity information, source 2... .
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| BT=NPDV |I|tag=1|pdvtyp | block length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
. ...PDV information for source 2... .
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| BT=NDEL |I|tag=1| resv | block length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
. ...delay information for source 2... .
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
This approach of separating the identification information is more costly than providing identification in each metrics block if only a single metrics block is sent in an RTCP packet, but becomes beneficial as soon as more than one metrics block shares common identification.
[Editor's note: the RTCP XR block namespace is limited by the 8-bit block type field in the RTCP XR header (Section 3 of [RFC3611]). IESG have noted that this is potentially restrictive. It would be possible to standardise an expansion mechanism, probably based on use of a new field near the start of the variable-length "type-specific block contents" field. Clearly this could apply only to new block types, so might be standardised to apply to some subrange of the current 8-bit range, for example the range 128 through 191 might be used. At time of writing, block types 12 to 254 are unassigned and 255 is reserved for future expansion. Is there a consensus for, or against, work to allow expansion? One potential use is through hierarchical control, where one or a few codepoints at the top level are given to other SDOs who may then define a number of metrics distinguished by values in the (so far hypothetical) new field.]
This section uses the example of an existing proposed metrics block to illustrate the application of the principles set out in Section 7.1.
The example [PDV] (work in progress) is a block to convey information about packet delay variation (PDV) only, consistent with the principle that a metrics block should address only one parameter of interest. One simple metric of PDV is available in the RTCP RR packet as the "jit" field. There are other PDV metrics which may be more useful to certain applications. Two such metrics are the IPDV metric ([Y1540], [RFC3393]) and the MAPDV2 metric [G1020]. Use of these metrics is consistent with the principle in Section 5 of [RFC5968] that metrics should usually be defined elsewhere, so that RTCP standards define only the transport of the metric rather than its nature. The purpose of this section is to illustrate the architecure using the example of [PDV] (work in progress) rather than to document the design of the PDV metrics block or to provide a tutorial on PDV in general.
Given the availability of at least three metrics for PDV, there are design options for the allocation of metrics to RTCP XR blocks:
In choosing between these options, extensibility is important, because additional metrics of PDV may well be standardised and require inclusion in this framework. The first option is extensible but only by use of additional RTCP XR blocks, which may consume the limited namespace for RTCP XR blocks at an unacceptable rate. The second option is not extensible, so could be rejected on that basis, but in any case a single application is quite unlikely to require transport of more than one metric for PDV. Hence the third option was chosen. This implies the creation of a subsidiary namespace to enumerate the PDV metrics which may be transported by this block, as discussed further in [PDV] (work in progress).
An RTP system (end system, mixer or translator) which originates, terminates or forwards RTCP XR blocks is expected to handle RTCP, including RTCP XR, as specified in [RFC3550] for that class of RTP systems. Provided this expectation is met, an RTP system using RTCP XR is architecturally no different from an RTP system of the same class (end system, mixer, or translator) which does not use RTCP XR. This statement applies to the topologies investigated in [RFC5117], where they use RTP end systems, RTP mixers and RTP translators as these classes are defined in [RFC3550].
These topologies are specifically Topo-Point-to-Point, Topo-Multicast, Topo-Translator (both variants, Topo-Trn-Translator and Topo-Media-Translator, and combinations of the two), and Topo-Mixer.
The topologies based on systems which do not behave according to [RFC3550], that is Topo-Video-Switch-MCU and Topo-RTCP-terminating-MCU, suffer from the difficulties described in [RFC5117]. These difficulties apply to systems sending, and expecting to receive, RTCP XR blocks as much as to systems using other RTCP packet types. For example, a participant RTP end system may send media to a video switch MCU. If the media stream is not selected for forwarding by the switch, neither RTCP RR packets nor RTCP XR blocks referring to the end system's generated stream will be received at the RTP end system. Strictly the RTP end system can only conclude that its RTP has been lost in the network, though an RTP end system complying with the robustness principle of [RFC1122] should survive with essential functions unimpaired.
Section 7.2 of [RFC3550] describes processing of RTCP by translators. RTCP XR is within the scope of the recommendations of [RFC3550]. Some RTCP XR metrics blocks may usefully be measured at, and reported by, translators. As described in [RFC3550] this creates a requirement for the translator to allocate an SSRC for itself so that it may populate the SSRC in the RTCP XR packet header (although the translator is not a Synchronisation Source in the sense of originating RTP media packets). It must also supply this SSRC and the corresponding CNAME in RTCP SDES packets.
In RTP sessions where one or more translators generate any RTCP traffic towards their next-neighbour RTP system, other translators in the session have a choice as to whether they forward a translator's RTCP packets. Forwarding may provide additional information to other RTP systems in the connection but increases RTCP bandwidth and may in some cases present a security risk. RTP translators may have forwarding behaviour based on local policy, which might differ between different interfaces of the same translator.
[Editor's note: for bidirectional unicast, an RTP system may usually detect RTCP from a translator by noting that the sending SSRC is not present in any RTP media packet. However even for bidirectional unicast there is a possibility of a source sending RTCP before it has sent any RTP media (leading to transient mis-categorisation of an RTP end system or RTP mixer as a translator), and for multicast sessions - or unidirectional/streaming unicast - there is a possibility of a receive-only end system being permanently mis-categorised as a translator. Is there a need for a translator to declare itself explicitly? Needs further thought.]
None.
This document itself contains no normative text and hence should not give rise to any new security considerations, to be confirmed.
The authors would like to thank Colin Perkins, Graeme Gibbs, Debbie Greenstreet, Keith Drage,Dan Romascanu, Ali C. Begen, Roni Even for their valuable comments and suggestions on the early version of this document.
The following are the major changes compared to previous version 00:
The following are the major changes compared to previous version 00: