Duplication Delay Attribute in the Session Description Protocol

A. Begen, Yujun Cai, Huang‐Tz Ou2014 · DOI

Abstract A straightforward approach to provide protection against packet losses due to network outages with a longest duration of T time units is to duplicate the original packets and send each copy separated in time by at least T time units. This approach is commonly referred to as "time-shifted redundancy", "temporal redundancy", or simply "delayed duplication". This document defines an attribute to indicate the presence of temporally redundant media streams and the duplication delay in the Session Description Protocol.

RFC 7197 Header Information

Introduction

In modern IP networks, packet delivery is frequently interrupted by hardware failures or link outages. While networks are designed with resiliency in mind, the process of "reconvergence"—the time required for the network to calculate and switch to a new path—typically spans tens to hundreds of milliseconds. During this critical window, packets are often dropped, leading to significant quality degradation in real-time media.

To mitigate this, engineers employ temporal redundancy (delayed duplication). By transmitting a second copy of a packet after a specific delay $T$, the system ensures that if the first copy is lost during a short outage, the second copy—arriving after the outage has likely resolved—will successfully reach the destination. However, for a receiver to effectively manage its buffers and suppress duplicate packets, it must know the exact duration of that delay. RFC 7197 formalizes the duplication-delay attribute within the Session Description Protocol (SDP) to provide this necessary signaling.

The duplication-delay Attribute

The core contribution of this protocol is the definition of a new declarative attribute: a=duplication-delay.

Syntax and Scope

Using Augmented Backus-Naur Form (ABNF), the attribute is defined as a=duplication-delay: periods, where "periods" consists of one or more integer values representing time in milliseconds. The attribute can be applied at two levels:

  • Session Level: Applying the delay specification to all streams within a session.
  • Media Level: Applying the specification to a specific media description.

Integration with SSRC Grouping

The attribute is designed to work in tandem with the "DUP" grouping semantics (defined in RFC 7104). This allows the SDP to group multiple Synchronization Sources (SSRCs) and then use the duplication-delay attribute to specify the temporal spacing between them. For instance, if a stream is sent three times (the original plus two copies), the attribute can list multiple values (e.g., a=duplication-delay: 50 100), indicating the delay of each subsequent copy relative to the original transmission.

Key Insights and Operational Logic

Simplicity and Use Cases

Unlike Forward Error Correction (FEC), which involves complex mathematical encoding and decoding, delayed duplication is a "buffer and resend" strategy. This makes it significantly easier to implement. It is particularly valuable in core IP networks where bandwidth is plentiful but "zero loss" is a high priority, such as high-quality real-time broadcast video feeds.

Duplicate Suppression

A critical requirement for the receiver is the ability to identify and discard duplicates to avoid audio/video artifacts. For RTP streams, this is handled via sequence numbers. For non-RTP applications, the receiver must utilize other unique identifiers to ensure the final output stream remains duplicate-free.

Congestion Warning

The RFC explicitly warns that duplication should not be used to address losses caused by network congestion. Because duplication doubles (or triples) the bandwidth requirements, applying it to a congested link will exacerbate the problem, leading to further packet loss for all traffic on that link.

Security and Limitations

  • Bandwidth Overhead: The primary trade-off for increased reliability is a 100% or greater increase in bandwidth consumption.
  • Security Risks: A malicious actor could modify the SDP to misinform a receiver about delays, potentially causing unnecessary buffering or packet loss. The authors recommend integrity protection via S/MIME or TLS for the signaling channel.
  • Denial of Service (DoS): There is a risk that attackers could signal an excessive number of duplicates to overwhelm a receiver's memory. Implementations are advised to set hard limits on the number of allowed duplicate streams.
  • Implementation Flexibility: While the RFC defines the protocol for signaling the delay, it does not mandate a specific algorithm for merging the streams; this remains at the discretion of the application developer.

Conclusion

RFC 7197 provides a standardized mechanism for signaling temporal redundancy in SDP. By enabling explicit communication of duplication delays, it allows receivers to more effectively utilize time-shifted streams to bridge the gap during network reconvergence, ensuring higher reliability for critical real-time media applications.

Summary updated

Abstract

A straightforward approach to provide protection against packet losses due to network outages with a longest duration of T time units is to duplicate the original packets and send each copy separated in time by at least T time units.This approach is commonly referred to as "time-shifted redundancy", "temporal redundancy", or simply "delayed duplication".This document defines an attribute to indicate the presence of temporally redundant media streams and the duplication delay in the Session Description Protocol.
BibTeX
@misc{begen2014duplication,
  author = {A. Begen and Yujun Cai and Huang‐Tz Ou},
  title = {{Duplication Delay Attribute in the Session Description Protocol}},
  year = {2014},
  doi = {10.17487/rfc7197},
}

Loading related content...

Figure