Indication of Load-balancing Strategy

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 13 April 2026.¶

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.¶

Table of Contents

1. Introduction

The load balancing serves as a critical mechanism for achieving high-performance transmission in wide-area networks (WANs), with diverse implementation strategies such as flow-level and packet-level load balancing, some approaches include:¶

*Equal-Cost Multi-Path (ECMP) and Weighted ECMP (WCMP): these mechanisms preserve the transmission order of packets within a data stream. When the network guarantees transmission and link reliability, packets typically arrive at the receiver in sequence. This eliminates the need for complex packet reordering at the receiver end, allowing simplified Go-Back-N (GBN)retransmission mechanisms to be adopted, thereby reducing device complexity and cost.¶

*Flowlet-based Load Balancing: this mechanism segments packets into micro-flowlets, which may be divided based on time intervals, message granularity, or fixed lengths. While these fragments are transmitted across different paths to maintain intra-flowlet ordering, they may cause inter-flowlet disorder, significantly increasing the receiver's buffer and reordering requirements.¶

*Packet Spraying: by randomly or load-aware proportionally distributing packets across multiple paths, this mechanism maximizes load balancing efficiency. However, it results in the highest packet disorder level, imposing the most stringent demands on the receiver's reordering capabilities.¶

In WANs with dynamic topologies, the load balancing strategy requires host-to-network collaboration to adapt to varying transmission scenarios. The selection of load balancing strategies should consider:¶

*traffic characteristics (e.g., flow size, burstiness, sensitivity to delay)¶

*path characteristics (e.g., latency, bandwidth, loss rate, and reliability)¶

*receiver's reordering capability (e.g., buffer size, reordering algorithm efficiency)¶

This document proposes the encoding of the indication for load-balancing strategies which can be encapsulated into a variety of protocols such as MPLS, IPv6 and SRv6 networks. It also provides the considerations for load-balancing configuration in control plane as well.¶

2. Conventions Used in This Document

3. Load Balancing Strategies

The reordering capabilities of incoming traffic flows may differ across servers. For example, standard RDMA NICs (Network Interface Cards) almost lack out-of-order reordering capabilities, while advanced DPU NICs achieve reordering through high-speed caches. The TCP receiving end can also perform out-of-order reordering based on caching. At network nodes, load balancing for flows is a local behavior and cannot perceive whether the server has the corresponding out-of-order recovery capability. The information from host-and-network collaboration may help to select the best load-balancing strategy.¶

          Traffic 1/2/3            Apply the load-balancing strategy
+--------+        +---------+      +-------+     +--------+      +---+---+     +-------+
|Client A|------->|Edge Node|----->|Node X |---->| Node A |----->|Node Y |<--->|Server |
+--------+        +---------+      +-------+  |  +--------+  ^   +-------+     +-------+
                 Insert the indication        |  +--------+  |
                 of load-balancing strategies +->| Node B |--+
                                              |  +--------+  ^
                                              |  +--------+  |
                                              +->| Node C |--+
                                                 +--------+
    Figure 1: Selection of Load-balancing Strategies

As shown in Figure 1, at the edge node, the information can be carried in packets to indicate the preferred load-balancing strategies. The indication may be carried in data packets throughout the network path. And the network node can route the packets based on the indication of load-balancing strategy, allowing the network to apply different strategies for different types of flows.¶

4. Encapsulation for Load-balancing Indication

    0 1 2 3 4 5 6 7
   +-+-+-+-+-+-+-+-+
   |  Flag   |F|L|P|
   +---------------+

    Figure 2: Indication Flag of Load-balancing Strategies

5. Considerations for Load-balancing Policy in Control Plane

The client or the server may notify the out-of-order recovery capability to the edge node or the controller in control plane. The controller can also sense the characteristics of a flow and decide the best load-balancing strategy and configure the indication to the edge node. And the edge node can encode the information into the encapsulation of the packets.¶

6. Security Considerations

To be discussed in future versions of this document.¶

7. IANA Considerations

TBA.¶

8. References

Authors' Addresses