Cloud-Scale RoCE Encapsulation for Multi-Tenant Layer-3 Networks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing RoCE protocols are limited by Layer-2 networks that do not scale and are not performant, lacking key properties of Layer-3 networks, leading to inefficiencies in data transfer and congestion management in public cloud environments.

Innovation Solution

Implementing techniques that enable Layer-2 RDMA traffic to be communicated over Layer-3 networks using Layer-3 protocols, with VLAN and QoS information preservation and congestion signaling, allowing for class-based queuing and separation across a shared network fabric.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If Layer-2 networks are used for RoCE traffic, then traffic isolation and QoS management are simplified, but scalability and network performance are limited

Engineering Contradiction:
ImprovescalabilityVSAvoidnetwork architecture complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent introduces a tunneling mechanism that encapsulates Layer-2 RoCE packets within Layer-3 IP packets. This intermediary approach allows RoCE traffic to traverse Layer-3 network infrastructure while preserving Layer-2 semantics, thereby achieving scalability without sacrificing traffic isolation or QoS management capabilities

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transitions from a pure Layer-2 network architecture to a layered approach where Layer-2 RoCE traffic is embedded within Layer-3 IP packets. This dimensional change enables the network to leverage Layer-3 routing and scaling capabilities while maintaining Layer-2 performance characteristics for RDMA operations

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Reliability

If dedicated VLANs are used for each tenant, then traffic isolation is maintained, but network resource utilization and scalability are reduced

Engineering Contradiction:
Improvetraffic isolationVSAvoidnetwork resource utilization
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent merges multiple tenant VLANs into a single shared Layer-3 network by encapsulating Layer-2 packets with Layer-3 IP headers that include tenant identification. This allows multiple tenants to share the same physical network infrastructure while maintaining logical isolation through IP-based routing and QoS policies

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The Layer-3 network infrastructure is designed to serve multiple functions simultaneously: it provides routing between tenants, enforces QoS policies, maintains traffic isolation through IP addressing, and supports RoCE traffic. This universal approach replaces the need for separate dedicated VLANs for each tenant

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Speed

If Layer-3 routing protocols are used for RDMA traffic, then network scalability and performance are improved, but traffic isolation and QoS management become more complex

Engineering Contradiction:
Improvedata transfer performanceVSAvoidQoS management complexity
Core Design Contradiction:
SpeedVSDevice complexity

Solution Approach 1:

The patent applies QoS markings and tenant identification to packets at the ingress point before they enter the Layer-3 network. This preliminary action ensures that QoS policies are established upfront, simplifying downstream packet handling and avoiding complex real-time QoS decisions throughout the network path

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4272408B1Cloud scale multi-tenancy for RDMA over converged ethernet (ROCE)
Publication Date: 2025.07.09 ORACLE INT CORP
  • EP4272408B1 patent drawingFigure 1
  • EP4272408B1 patent drawingFigure 2
  • EP4272408B1 patent drawingFigure 3

AI summary

Techniques and apparatus for data networking are described. In one example, a method includes receiving a first Layer-2 Remote Direct Memory Access (RDMA) packet which includes a virtual local area network (VLAN) tag and a quality-of-service (QoS) data field; converting the first Layer-2 RDMA packet to a first Layer-3 encapsulated packet; and forwarding the first Layer-3 encapsulated packet to a switch fabric. In this method, the converting includes adding at least one header to the first Layer-2 RDMA packet, where the at least one header includes: a virtual network identifier that is based on information from the VLAN tag, and a QoS value that is based on information from the QoS data field.