| Network Working Group | Z. Sarker |
| Internet-Draft | Ericsson AB |
| Intended status: Informational | V. Singh |
| Expires: December 19, 2014 | Aalto University |
| X. Zhu | |
| M. Ramalho | |
| Cisco Systems | |
| June 17, 2014 |
Test Cases for Evaluating RMCAT Proposals
draft-sarker-rmcat-eval-test-01
The Real-time Transport Protocol (RTP) is used to transmit media in multimedia telephony applications, these applications are typically required to implement congestion control. The RMCAT working group is currently working on candidate algorithms for such interactive real-time multimedia applications. This document describes the test cases to be used in the performance evaluation of those candidate algorithms.
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 December 19, 2014.
Copyright (c) 2014 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.
This memo describes a set of test cases for evaluating candidate RMCAT congestion control algorithm proposals, it is based on the guidelines enumerated in [I-D.ietf-rmcat-eval-criteria] and the requirements discussed in [I-D.ietf-rmcat-cc-requirements]. The test cases cover basic usage scenarios and are described using a common structure, which allows for additional test cases to be added to those described herein to accommodate other topologies and/or the modeling of different path characteristics. It is the intention of this work to capture the consensus of the RMCAT working group participants regarding the test cases upon which the performance of the candidate RMCAT proposals should be evaluated.
The terminology defined in RTP [RFC3550], RTP Profile for Audio and Video Conferences with Minimal Control [RFC3551], RTCP Extended Report (XR) [RFC3611], Extended RTP Profile for RTCP-based Feedback (RTP/AVPF) [RFC4585], Support for Reduced-Size RTCP [RFC5506], and RTP Circuit Breaker algorithm [I-D.ietf-avtcore-rtp-circuit-breakers] apply.
All test cases in this document follow a basic structure allowing implementers to describe a new test scenario without repeatedly explaining common attributes. The structure includes a general description section that describes the test case and its motivation. Additionally the test case defines a set of attributes that characterize the testbed, i.e., the network path between communicating peers and the diverse traffic sources.
+---+ +---+
|S1 |====== \ Forward --> / =======|R1 |
+---+ \\ // +---+
\\ //
+---+ +-----+ +-----+ +---+
|S2 |=======| A |------------------------------>| B |=======|R2 |
+---+ | |<------------------------------| | +---+
+-----+ +-----+
(...) // \\ (...)
// <-- Backward \\
+---+ // \\ +---+
|Sn |====== / \ ======|Rn |
+---+ +---+
Figure 1: Example of A Testbed Topology
Any attribute can have a set of values (enclosed within "[]"). Each member value of such a set MUST be treated as different value for the same attribute. It is desired to run separate tests for each such attribute value.
The test cases described in this document follow the above structure.
This section describes recommended test case settings and could be overwritten by the respective test cases.
To evaluate the performance of the candidate algorithms it is expected to log enough information to visualize the following metrics at a fine enough time granularity:
Each path between a sender and receiver as described in Figure 1 have the following characteristics unless otherwise specified in the test case.
Examples of additional network parameters are discussed in Appendix A.
Unless otherwise specified, each test case will include one or more media sources as described below.
In this test case the bottleneck-link capacity between the two endpoints varies over time. This test is designed to measure the responsiveness of the candidate algorithm. This test tries to address the requirements in [I-D.ietf-rmcat-cc-requirements], which requires the algorithm to adapt the flow(s) and provide lower end-to-end latency when there exists:
It should be noted that the exact variation in available capacity due to any of the above depends on the under-lying technologies. Hence, we describe a set of known factors, which may be extended to devise a more specific test case targeting certain behaviour in a certain network environment.
Expected behavior: the candidate algorithm is expected to detect the path capacity constraint, converges to bottleneck link's capacity and adapt the flow to avoid unwanted oscillation when the sending bit rate is approaching the bottleneck link's capacity. The oscillations occur when the media flow(s) attempts to reach its maximum bit rate, overshoots the usage of the available bottleneck capacity, to rectify it reduces the bit rate and starts to ramp up again.
Forward -->
+---+ +-----+ +-----+ +---+
|S1 |=======| A |------------------------------>| B |=======|R1 |
+---+ | |<------------------------------| | +---+
+-----+ +-----+
<-- Backward
Figure 2: Testbed Topology for Limited Link Capacity
Testbed topology: One media source S1 is connected to corresponding R1. The media traffic is transported over the forward path and corresponding feedback/control traffic is transported over the backward path.
To evaluate the performance of the candidate algorithms it is expected to log enough information to visualize the metrics described in Section 4.1 at a fine enough time granularity.
Testbed attributes:
| Variation pattern index | Path direction | Start time | Path capacity |
|---|---|---|---|
| One | Forward | 0s | 1Mbps |
| Two | Forward | 40s | 2.5Mbps |
| Three | Forward | 60s | 600Kbps |
| Four | Forward | 80s | 1Mbps |
This test case is similar to Section 5.1. However in addition this test will also consider persistent network load due to competing traffic.
Expected behavior: the candidate algorithms is expected to detect the variation in available capacity and adapt the media stream(s) accordingly. The flows stabilize around their maximum bitrate as the as the maximum link capacity is large enough to accommodate the flows. When the available capacity drops, the flow(s) adapts by decreasing its sending bit rate, and when congestion disappears, the flow(s) are again expected to ramp up.
To evaluate the performance of the candidate algorithms it is expected to log enough information to visualize the metrics described in Section 4.1 at a fine enough time granularity:
+---+ +---+
|S1 |===== \ / =======|R1 |
+---+ \\ Forward --> // +---+
\\ //
+-----+ +-----+
| A |------------------------------>| B |
| |<------------------------------| |
+-----+ +-----+
// \\
// <-- Backward \\
+---+ // \\ +---+
|S2 |====== / \ ======|R2 |
+---+ +---+
Figure 3: Testbed Topology for Variable Available Capacity
Testbed Topology: Two (2) media sources S1 and S2 are connected to their corresponding destinations R1 and R2. The media traffic is transported over the forward path and corresponding feedback/control traffic is transported over the backward path.
Testbed attributes:
Testbed attributes are similar as described in Section 5.1 except the test specific capacity variation setup.
Test Specific Information: This test uses path capacity variation as listed in Table 2 with a corresponding end time of 125 seconds.
| Variation pattern index | Path direction | Start time | Path capacity |
|---|---|---|---|
| One | Forward | 0s | 4Mbps |
| Two | Forward | 25s | 2Mbps |
| Three | Forward | 50s | 3.5Mbps |
| Four | Forward | 75s | 1Mbps |
| Five | Forward | 100s | 2Mbps |
RMCAT WG has been chartered to define algorithms for RTP hence it is assumed that RTCP, RTP header extension or such would be used by the congestion control algorithm in the backchannel. Due to asymmetric nature of the link between communicating peers it is possible for a participating peer to not receive such feedback information due to an impaired or congested backchannel (even when the forward channel might not be impaired). This test case is designed to observe the candidate congestion control behaviour in such an event.
It is expected that the candidate algorithms is able to cope with the lack of feedback information and adapt to minimize the performance degradation of media flows in the forward channel.
It should be noted that for this test case: logs are compared with the reference case, i.e, when the backward channel has no impairments
To evaluate the performance of the candidate algorithms it is expected to log enough information to visualize the metrics described in Section 4.1 at a fine-grained time intervals:
Testbed topology: One (1) media source S1 is connected to corresponding R1, but both endpoints are additionally receiving and sending data, respectively. The media traffic (S1->R1) is transported over the forward path and corresponding feedback/control traffic is transported over the backward path. Likewise media traffic (S2->R2) is transported over the backward path and corresponding feedback/control traffic is transported over the forward path.
+---+ +---+
|S1 |===== \ Forward --> / =======|R1 |
+---+ \\ // +---+
\\ //
+-----+ +-----+
| A |------------------------------>| B |
| |<------------------------------| |
+-----+ +-----+
// \\
// <-- Backward \\
+---+ // \\ +---+
|R2 |===== / \ ======|S2 |
+---+ +---+
Figure 4: Testbed Topology for Congested Feedback Link
Testbed attributes:
| Variation pattern index | Path direction | Start time | Path capacity |
|---|---|---|---|
| One | Forward | 0s | 2Mbps |
| Two | Forward | 20s | 1Mbps |
| Three | Forward | 40s | 500Kbps |
| Four | Forward | 60s | 2Mbps |
| Variation pattern index | Path direction | Start time | Path Capacity |
|---|---|---|---|
| One | Backward | 0s | 2Mbps |
| Two | Backward | 35s | 800Kbps |
| Three | Backward | 70s | 2Mbps |
In this test case, more than one RMCAT media flow shares the bottleneck link and each of them uses the same congestion control algorithm. This is a typical scenario where a real-time interactive application sends more than one media flows to the same destination and these flows are multiplexed over the same port. In such a scenario it is likely that the flows will be routed via the same path and need to share the available bandwidth amongst themselves. For the sake of simplicity it is assumed that there are no other non-RMCAT competing traffic sources in the bottleneck link and that there is sufficient capacity to accommodate all the flows individually. While this appears to be a variant of the test case defined in Section 5.2, it focuses on the capacity sharing aspect of the candidate algorithm. The previous test case, on the other hand, measures adaptability, stability, and responsiveness of the candidate algorithm.
Expected behavior: It is expected that the competing flows will converge to an optimum bit rate to accommodate all the flows with minimum possible latency and loss. Specifically, the test introduces three media flows at different time instances, when the second flow appears there should still be room to accommodate another flow on the bottleneck link. Lastly, when the third flow appears the bottleneck link should be saturated.
To evaluate the performance of the candidate algorithms it is expected to log enough information to visualize the metrics described in Section 4.1 at a fine enough time granularity:
+---+ +---+
|S1 |===== \ Forward --> / =======|R1 |
+---+ \\ // +---+
\\ //
+---+ +-----+ +-----+ +---+
|S2 |=======| A |------------------------------>| B |=======|R2 |
+---+ | |<------------------------------| | +---+
+-----+ +-----+
// \\
// <-- Backward \\
+---+ // \\ +---+
|S3 |====== / \ ======|R3 |
+---+ +---+
Figure 5: Testbed Topology for Multiple RMCAT Flows
Testbed topology: Three media sources S1, S2, S3 are connected to respective R1, R2, R3. The media traffic is transported over the forward path and corresponding feedback/control traffic is transported over the backward path.
Testbed attributes:
In this test case, multiple RMCAT media flows share the bottleneck link, but the end-to-end path latency for each RMCAT flow is different. For the sake of simplicity it is assumed that there are no other non-RMCAT competing traffic sources in the bottleneck link and that there is sufficient capacity to accommodate all the flows. While this appears to be a variant of test case 5.2, it focuses on the capacity sharing aspect of the candidate algorithm under different RTTs.
It is expected that the competing flows will converge to bit rates to accommodate all the flows with minimum possible latency and loss. Specifically, the test introduces five media flows at the same time instance.
To evaluate the performance of the candidate algorithms it is expected to log enough information to visualize the metrics described in Section 4.1 at a fine enough time granularity:
Testbed Topology: Five (5) media sources S1,S2,..,S5 are connected to their corresponding media sinks R1,R2,..,R5. The media traffic is transported over the forward path and corresponding feedback/control traffic is transported over the backward path. The topology is the same as in Section 5.4. The end-to-end path delays are: 10ms for S1-R1, 25ms for S2-R2, 50ms for S3-R3, 100ms for S4-R4, and 150ms S5-R5, respectively.
Testbed attributes:
In this test case, one or more RMCAT media flows share the bottleneck link with at least one long lived TCP flows. Long lived TCP flows download data throughout the session and are expected to have infinite amount of data to send and receive. This is a scenario where a multimedia application co-exists with a large file download. The test case measures the adaptivity of the candidate algorithm to competing traffic. It addresses the requirement 3 in [I-D.ietf-rmcat-cc-requirements].
Expected behavior: depending on the convergence observed in test case 5.1 and 5.2, the candidate algorithm may be able to avoid congestion collapse. In the worst case, the media stream will fall to the minimum media bit rate.
To evaluate the performance of the candidate algorithms it is expected to log enough information to visualize the following metrics in addition to the metrics described in Section 4.1 at a fine enough time granularity:
Testbed topology: One (1) media source S1 is connected to corresponding media sink, R1. In addition, there is a long-live TCP flow sharing the same bottleneck link. The media traffic is transported over the forward path and corresponding feedback/control traffic is transported over the backward path. The TCP traffic goes over the forward path from, S_tcp with acknowledgement packets flowing along the backward path from, R_tcp.
+--+ +--+
|S1|===== \ Forward --> / =======|R1|
+--+ \\ // +--+
\\ //
+-----+ +-----+
| A |---------------------------->| B |
| |<----------------------------| |
+-----+ +-----+
// \\
// <-- Backward \\
+-----+ // \\ +-----+
|S_tcp|=== / \ ===|R_tcp|
+-----+ +-----+
Figure 6: Testbed Topology for TCP vs RMCAT Flows
Testbed attributes:
In this test case, one or more RMCAT media flow shares the bottleneck link with multiple short-lived TCP flows. Short-lived TCP flows resemble the on/off pattern observed in the web traffic, wherein clients (browsers) connect to a server and download a resource (typically a web page, few images, text files, etc.) using several TCP connections (up to 4). This scenario shows the performance of the multimedia application when several browser windows are active. The test case measures the adaptivity of the candidate algorithm to competing web traffic, it addresses the requirements 1.E in [I-D.ietf-rmcat-cc-requirements].
Depending on the number of short TCP flows, the cross-traffic either appears as a short burst flow or resembles a long TCP flow. The intention of this test is to observe the impact of short-term burst on the behaviour of the candidate algorithm.
To evaluate the performance of the candidate algorithms it is expected to log enough information to visualize the following metrics in addition to the metrics described in Section 4.1 at a fine enough time granularity:
Testbed topology: The topology described here is same as the one described in Figure 6.
Testbed attributes:
In this test case, more than one real-time interactive media flows share the link bandwidth and all flows reach to a steady state by utilizing the link capacity in an optimum way. At these stage one of the media flow is paused for a moment. This event will result in more available bandwidth for the rest of the flows and as they are on a shared link. When the paused media flow will resume it would no longer have the same bandwidth share on the link. It has to make it's way through the other existing flows in the link to achieve a fair share of the link capacity. This test case is important specially for real-time interactive media which consists of more than one media flows and can pause/resume media flow at any point of time during the session. This test case directly addresses the requirement number 5 in [I-D.ietf-rmcat-cc-requirements]. One can think it as a variation of test case defined in Section 5.4. However, it is different as the candidate algorithms can use different strategies to increase its efficiency, for example the fairness, convergence time, reduce oscillation etc, by capitalizing the fact that they have previous information of the link.
To evaluate the performance of the candidate algorithms it is expected to log enough information to visualize the following metrics in addition to the metrics described in Section 4.1 at a fine enough time granularity:
Testbed Topology: Same as test case defined in Section 5.4
Testbed attributes: The general description of the testbed parameters are same as Section 5.4 with changes in the test specific setup as below-
It has been noticed that there are other interesting test cases besides the basis test cases listed above. In many aspects, these additional test cases can help to further evaluate the candidate algorithm. They are listed as below.
This test case requires to run all the basic test cases with the availability of Explicit Congestion Notification (ECN) [RFC6679] feature enabled. The goal of this test is to exhibit that the candidate algorithms does not fail when ECN signals are available. With ECN signals enabled the algorithms are expected to perform better than their delay based variants.
In this test case one RMCAT flow, S1->R2 traverse a path with multiple bottlenecks. As illustrated in Figure 7, the first flow (S1->R1) competes with the second RMCAT flow (S2->R2) over the link between A and B which is close to the sender side; again, that flow (S1->R1) competes with the third RMCAT flow (S3->R3) over the link between C and D which is close to the receiver side. The goal of this test is to ensure that the candidate algorithms work properly in the presence of multiple bottleneck links on the end to end path.
Expected behavior: the candidate algorithm is expected to achieve full utilization at both bottleneck links without starving any of the three RMCAT flows.
Forward ---->
+---+ +---+ +---+ +---+
|S2 | |R2 | |S3 | |R3 |
+---+ +---+ +---+ +---+
| | | |
| | | |
+---+ +-----+ +-----+ +-----+ +-----+ +---+
|S1 |=======| A |------>| B |----->| C |---->| D |=======|R1 |
+---+ | |<------| |<-----| |<----| | +---+
+-----+ +-----+ +-----+ +-----+
1st 2nd
Bottleneck (A->B) Bottleneck (C->D)
<------ Backward
Figure 7: Testbed Topology for Multiple Bottlenecks
Testbed topology: Three media sources S1, S2, and S3 are connected to respective destinations R1, R2, and R3. For all three flows the media traffic is transported over the forward path and corresponding feedback/control traffic is transported over the backward path.
Testbed attributes:
Additional cellular network specific test cases are define in [I-D.draft-sarker-rmcat-cellular-eval-test-cases]
TBD
[Editor's Note: We should encourage people to come up with possible WiFi Network specific test cases]
Security issues have not been discussed in this memo.
There are no IANA impacts in this memo.
Much of this document is derived from previous work on congestion control at the IETF.
The content and concepts within this document are a product of the discussion carried out in the Design Team.
In addition to the recommended evaluation settings in Section 4, the implemntors can extend their tests by choosing from the following values:
Experiments are expected to verify that the congestion control is able to work in challenging situations, for example over trans-continental and/or satellite links. Typical values are:
To model lossy links, the experiments can choose one of the following loss rates, the fractional loss is the ratio of packets lost and packets sent.
The router queue length is measured as the time taken to drain the FIFO queue. It has been noted in various discussions that the queue length in the current deployed Internet varies significantly. While the core backbone network has very short queue length, the home gateways usually have larger queue length. Those various queue lengths can be categorized in the following way:
Here the size of the queue is measured in bytes or packets and to convert the queue length measured in seconds to queue length in bytes:
QueueSize (in bytes) = QueueSize (in sec) x Throughput (in bps)/8
This section defines jitter model for the purposes of this document. When jitter is to be applied to both the RMCAT flow and any competing flow (such as a TCP competing flow), the competing flow will use the jitter definition below that does not allow for re-ordering of packets on the competing flow (see NR-RBPDV definition below).
Jitter is an overloaded term in communications. Its meaning is typically associated with the variation of a metric (e.g., delay) with respect to some reference metric (e.g., average delay or minimum delay). For example, RFC 3550 jitter is a smoothed estimate of jitter which is particularly meaningful if the underlying packet delay variation was caused by a Gaussian random process.
Because jitter is an overloaded term, we instead use the term Packet Delay Variation (PDV) to describe the variation of delay of individual packets in the same sense as the IETF IPPM WG has defined PDV in their documents (e.g., RFC 3393) and as the ITU-T SG16 has defined IP Packet Delay Variation (IPDV) in their documents (e.g., Y.1540).
Most PDV distributions in packet network systems are one-sided distributions (the measurement of which with a finite number of measurement samples result in one-sided histograms). In the usual packet network transport case there is typically one packet that transited the network with the minimum delay, then a majority of packets also transit the system within some variation from this minimum delay, and then a minority of the packets transits the network with delays higher than the median or average transit time (these are outliers). Although infrequent, outliers can cause significant deleterious operation in adaptive systems and should be considered in RMCAT adaptation designs.
In this section we define two different bounded PDV characteristics, 1) Random Bounded PDV and 2) Approximately Random Subject to No-Reordering Bounded PDV.
Random Bounded PDV (RBPDV):
The RBPDV probability distribution function (pdf) is specified to be of some mathematically describable function which includes some practical minimum and maximum discrete values suitable for testing. For example, the minimum value, x_min, might be specified as the minimum transit time packet and the maximum value, x_max, might be idefined to be two standard deviations higher than the mean.
Since we are typically interested in the distribution relative to the mean delay packet, we define the zero mean PVD sample, z(n), to be z(n) = x(n) - x_mean, where x(n) is a sample of the RBPDV random variable x and x_mean is the mean of x.
We assume here that s(n) is the original source time of packet n and the post-jitter induced emmission time, j(n), for packet n is j(n) = {[z(n) + x_mean] + s(n)}. It follows that the separation in the post-jitter time of packets n and n+1 is {[s(n+1)-s(n)] - [z(n)-z(n+1)]}. Since the first term is always a positive quantity, we note that packet reordering at the receiver is possible whenever the second term is greater than the first. Said another way, whenever the difference in possible zero mean PDV sample delays (i.e., [x_max-x_min]) exceeds the inter-departure time of any two sent packets, we have the possibility of packet re-ordering.
There are important use cases in real networks where packets can become re-ordered such as in load balancing topologies and during route changes. However, for the vast majority of cases there is no packet re-ordering because most of the time packets follow the same path. Due to this, if a packet becomes overly delayed, the packets after it on that flow are also delayed. This is especially true for mobile wireless links where there are per-flow queues prior to base station scheduling. Owing to this important use case, we define another PDV profile similar to the above, but one that does not allow for re-ordering within a flow.
Approximately Random Subject to No-Reordering Bounded PDV (NR-RPVD):
No Reordering RPDV, NR-RPVD, is defined similarly to the above with one important exception. Let serial(n) be defined as the serialization delay of packet n at the lowest bottleneck link rate (or other appropriate rate) in a given test. Then we produce all the post-jitter values for j(n) for n = 1, 2, ... N, where N is the length of the source sequence s to be jittered. The exception can be stated as follows: We revisit all j(n) beginning from index n=2, and if j(n) is determined to be less than [j(n-1)+serial(n-1)], we redefine j(n) to be equal to [j(n-1)+serial(n-1)] and continue for all remaining n (i.e., n = 3, 4, .. N). This models the case where the packet n is sent immediately after packet (n-1) at the bottleneck link rate. Although this is generally the theoretical minimum in that it assumes that no other packets from other flows are in-between packet n and n+1 at the bottleneck link, it is a reasonable assumption for per flow queuing.
We note that this assumption holds for some important exception cases, such as packets immediately following outliers. There are a multitude of software controlled elements common on end-to-end Internet paths (such as firewalls, ALGs and other middleboxes) which stop processing packets while servicing other functions (e.g., garbage collection). Often these devices do not drop packets, but rather queue them for later processing and cause many of the outliers. Thus NR-RPVD models this particular use case (assuming serial(n+1) is defined appropriately for the device causing the outlier) and thus is believed to be important for adaptation development for RMCAT.
[Editor's Note: It may require to define test distributions as well. Example test distrubution may include-
1 - Two-sided: Uniform PDV Distribution. Two quantities to define: x_min and x_max.
2 - Two-sided: Truncated Gaussian PDV Distribution. Four quantities to define: the appropriate x_min and x_max for test (e.g., +/- two sigma values), the standard deviation and the mean.
3 - One Sided: TBD]
[Editor's note : Describes the model for generating packet losses, for example, losses can be generated using traces, or using the Gilbert-Elliot model, or randomly (uncorrelated loss).]
Long-lived TCP flows will download data throughout the session and are expected to have infinite amount of data to send or receive.
Each short TCP flow is modeled as a sequence of file downloads interleaved with idle periods. Not all short TCPs start at the same time, i.e., some start in the ON state while others start in the OFF state.
The short TCP flows can be modelled in two ways, 1) 100s of flows fetching small (5-20 KB) amounts of data, or 2) 10s of flows fetching slightly larger (100-1000KB) amounts of data.
The idle period is typically derived from an exponential distribution with the mean value of 10 seconds.
[Open issue: short-lived/bursty TCP cross-traffic parameters are still to be agreed upon].