SCONE Working Group D. Druta
Internet-Draft E. Halepovic
Intended status: Informational T. Karagioules
Expires: 3 August 2025 AT&T
30 January 2025
Video Session Data Rate for SCONE protocol
draft-druta-scone-video-session-data-rate-00
Abstract
The SCONE protocol requires a semantically consistent way for CSPs to
convey a throughput advice. The Video Session Data Rate (VSDR)
describes the formula to be applied both for setting the limit on the
CAP side as well as for the CSP to validate conformance with the
policy.
Status of This Memo
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 https://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 3 August 2025.
Copyright Notice
Copyright (c) 2025 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 (https://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 Revised BSD License text as
described in Section 4.e of the Trust Legal Provisions and are
provided without warranty as described in the Revised BSD License.
Druta, et al. Expires 3 August 2025 [Page 1]
Internet-Draft VSDR January 2025
Table of Contents
1. Motivation and Drivers . . . . . . . . . . . . . . . . . . . 2
2. Objectives . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 3
4. Details . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
5. Benefits and Uses . . . . . . . . . . . . . . . . . . . . . . 4
6. Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . 5
7. Limitations . . . . . . . . . . . . . . . . . . . . . . . . . 6
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . 6
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 6
1. Motivation and Drivers
The SCONE working group has started to work on a protocol allowing
CSPs (Communication Service Providers) to pass throughput advice to
CAPs (Content and Application Providers) for the purpose of enforcing
network policies associated with video throughput limits in a self-
regulatory fashion.
While the actual SCONE mechanism has not been defined yet, a simple,
standard semantic representation of the throughput advice is
necessary in order to have a common baseline measure to track overall
video session data consumption. This draft addresses the formula of
metric to be used for the SCONE protocol advise.
Since a significant percentage of the internet traffic consists of
video and specifically Adaptive Bitrate Video, the focus is on video
sessions and their impact on the subscription limits. For most, the
active stream being displayed on the screen is the only one
delivered. That being said, many video services use QUIC HTTP/3 as a
delivery mechanism and take advantage of the multiplexing
capabilities provided by the protocol. Therefore, at any given point
in time it is possible to have multiple video streams delivered over
the same connection.
In other words, the requirement is to establish a measurement that
represents the Aggregate Bitrate of all the active video flows
associated with a particular 4 tuple.
An additional requirement is to establish an easily computable metric
by both CSPs and CAPs that can be compared and measured by both
sides.
Druta, et al. Expires 3 August 2025 [Page 2]
Internet-Draft VSDR January 2025
2. Objectives
On the CSP side, since most of the internet traffic is end-to-end
encrypted, the main objective is to be able to characterize video
sessions from network data in the absence of application-layer
metrics.
In addition to that, there is a need to estimate customer experience
and network support for video delivery while assessing how is video
service using available bandwidth.
On the CAP side, using the same formula to compute the data usage for
video delivery allows for a proactive implementation of video
optimization delivery techniques in compliance and alignment with the
network policies set for the subscription.
3. Definitions
A video session, from the network perspective is the set of IP flows
carrying video (and audio and metadata, as appropriate), which are
separated from other such flows by a timeout T.
Data is the total volume of all flows in the session per direction of
transmission, typically downlink for streaming video, and both
downlink and uplink for video conferencing.
Duration is the period from the earliest data transmission and the
latest data transmission across all flows in the video session,
eliminating idle times at the beginning and end of the flows, if
exclusion is possible.
4. Details
Figure 1 below illustrates three video sessions, comprised of
different video streams (three active ones with some idle periods
representing a typical video delivery scenario with a main video
being displayed and others being queued or pre-fetched in the
background.
Druta, et al. Expires 3 August 2025 [Page 3]
Internet-Draft VSDR January 2025
Session A | Delta_t > T |Session B |Delta t > T Video Session C |
| | | | | |
|<--------->|<----------->|<-------->|<------>|<---|------|------->|
Flow #1 |XXXX | o|OOOOOOOOO| | | | ######|
Flow #2 | XXXXXX|XX / | OOO | | | |####### |
Flow #3 | XXXX | \ / | | |####| | |
|<--------->| \ / | | | |<---->| |
| | \ / | | | | | |
Duration D Idle period Delta t < T
(excluded)
Figure 1: Three flow traffic for Video Session Data Rate
The Video Session Data Rate or VSDR is ratio (expressed in Kbit/s or
Mbit/s) of Data (Cumulative data transmitted over all the video
flows) and duration D.
5. Benefits and Uses
There are several benefits to defining a standard VSDR. For once,
expressing the metric using the above-described formula provides
simplicity of expression and calculation. In addition to that, it
enables robustness against certain types of abuse.
Since this measure is an aggregate bitrate of active video flows, it
should be able to provide some correlation with encoding bitrate.
Additionally, VSDR is easily computable by both CAPs and CSPs and
therefore induces very little extra compute cycles in the process.
VSDR is robust to some forms of miscalculation. First, it reflects
the true overall data rate of the session compared to averaging or
otherwise summarizing flow throughputs. Therefore, it is robust to
outliers resulting from very short flows or flows with very little
data. Second, it is robust to some form of miscalculation. By
establishing a sensible timeout T, session concatenation is
discouraged for long periods of time resulting in artificially lower
VSDR values.
This method reflects more accurately the video delivery behavior and
is measurable based on IP traffic patterns observed on the wire.
Examples of computation:
Druta, et al. Expires 3 August 2025 [Page 4]
Internet-Draft VSDR January 2025
CSP can compute VSDR using high-level flow records that contain start
and end times, and data volume per direction. No packet-level data
is required. Therefore, a simple accumulation of flow records can
provide for quick computation of session start and end time for
duration, and data volume summation, while separating sessions based
on the timeout. This can be done in near real time as well as later.
CAP can compute the same VSDR using more than one method. At the CDN
edge note, flow monitoring towards different ASNs or subnets can
compute VSDR in the same way as the CSP. Web server logs can be used
as well where multiple request/response pairs will comprise each
flow.
VSDR can be used in the following ways:
For the purpose of SCONE, it provides a flexible and versatile
vehicle to convey throughput advice or a target rate: A CSP can use
VSDR to express to the CAP the maximum rate of delivery of the data.
Depending on CSP network or subscription policy, it can be used as
per-session, per-minute, or even as a percentile (e.g., 95% of
sessions or daily data should be delivered in sessions under the
certain VSDR). VSDR).
Quality of delivery estimation: As a proxy for overall delivery, its
increase or decrease can be used as a proxy for improvement or
degradation of the delivery of content. Slower sessions can
represent impact of network load and congestion, and reaction of a
video service to it.
6. Assumptions
In order to measure the VSDR, on the network side it is necessary to
correctly and accurately identify video flows. Since the majority of
the video traffic over the internet is end to end encrypted, the
identification of flows is commonly based on heuristics and other
techniques. Although CSPs can identify the video flows, applying
traffic shaping filters in the network can negatively impact the
Quality of Experience for the video service as there is no video
delivery optimization possible. By allowing the CAP to adjust and
optimize within the VSDR limit, the expectation is that not only the
network policies are enforced but the video streams are efficiently
distributed to provide users with a good user experience.
Druta, et al. Expires 3 August 2025 [Page 5]
Internet-Draft VSDR January 2025
7. Limitations
Since the VSDR is a measure relative to the individual CAP
connections, it is not always representative of customer experience.
That is because at any given time, especially on mobile connections,
congestion can be a factor. In addition to that, individual CAP
traffic cannot be completely isolated from the rest of the traffic on
the same device. A background download for a software update, for
example could have an impact on the overall connection throughput.
Changes in VSDR value are comparable within each video service across
network conditions under uniform use assumption, but not comparable
across different video services. For example, if we assume that a
user population behavior of one video service is uniform across
geographic and network locations (meaning that the distribution of
video bitrates and general usage are comparable), then differences in
VSDR can indicate difference in performance and QoE across network
locations, within that video service. However, due to differences in
user behavior, stream design and other factors, VSDR differences are
generally not comparable between different video services. An
example may be the use of the more efficient codec in one video
service vs another.
Acknowledgements
Thanks to Abhishek Tiwari for providing input and feedback.
Authors' Addresses
Dan Druta
AT&T
Email: dd5826@att.com
Emir Halepovic
AT&T
Email: emir@research.att.com
Theo Karagioules
AT&T
Email: theo@research.att.com
Druta, et al. Expires 3 August 2025 [Page 6]