RDMA NIC Hardware Offload for Multi-Tenant Overlay Encapsulation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current RDMA protocols, such as RoCE, do not support multi-tenancy in Software Defined Data Centers and incur significant CPU costs due to software handling of overlay network address mapping and encapsulation/decapsulation processes, affecting performance.

Innovation Solution

Implementing an overlay network that encapsulates RDMA data transfer protocols with overlay network routing headers, offloading encapsulation to hardware through programming the RDMA NIC with overlay network address mapping information, enabling L3 routability and multi-tenancy in cloud computing environments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If software handles overlay network address mapping and encapsulation/decapsulation processes, then multi-tenancy is supported, but CPU overhead increases significantly

Engineering Contradiction:

Improvemulti-tenancy supportVSAvoidCPU overhead

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent replaces software-based encapsulation/decapsulation with hardware-based implementation using RDMA NIC. The NIC is programmed with overlay network address mapping information and performs encapsulation/decapsulation operations in hardware, substituting the CPU-intensive software process with efficient hardware operations that reduce CPU overhead while maintaining multi-tenancy support

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The RDMA NIC is programmed with overlay network address mapping information specific to each RDMA connection, enabling it to perform encapsulation/decapsulation operations autonomously without CPU intervention. The hardware self-services the multi-tenancy requirements by maintaining and using its own address mapping tables, freeing the CPU from these operations

Inventive Principle:

Principle #25Self-service

2Productivity

If RDMA protocols are used over Ethernet networks, then direct data transfers between hosts are enabled, but multi-tenancy and L3 routing support are lacking

Engineering Contradiction:

Improvedata transfer speedVSAvoidmulti-tenancy support

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent implements nested encapsulation where RDMA packets are encapsulated within overlay network packets. The RDMA data transfer protocol packets are nested inside overlay network packets that contain L3 routing headers and overlay network identification information. This nested structure enables RDMA to operate over L3 networks while maintaining its high-speed direct transfer capability and adding multi-tenancy support through the outer overlay network headers

Inventive Principle:

Principle #7Nested doll (Nesting)

Solution Approach 2:

The overlay network protocol acts as an intermediary layer between RDMA and the underlying Ethernet network. It provides L3 routing capabilities and multi-tenancy support without interfering with the RDMA data transfer mechanism. The intermediary overlay network enables RDMA to function in multi-tenant environments with L3 routing while preserving the original high-speed direct transfer特性

Inventive Principle:

Principle #24Intermediary (Mediator)

3Adaptability or versatility

If overlay network encapsulation is performed in software, then multi-tenancy is achieved, but performance is degraded due to CPU costs

Engineering Contradiction:

Improvemulti-tenancy supportVSAvoidpacket processing speed

Core Design Contradiction:

Adaptability or versatilityVSSpeed

Solution Approach 1:

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent extracts the encapsulation/decapsulation function from the software domain and relocates it to hardware domain. By programming the RDMA NIC with overlay network address mapping information, the encapsulation/decapsulation operations are performed in hardware, removing the performance-degrading software processing step while preserving multi-tenancy functionality

Inventive Principle:

Principle #2Taking out (Extraction)

Data Source

PatentUS11782868B2Methods and systems to achieve multi-tenancy in RDMA over converged Ethernet

Publication Date: 2023.10.10 VMWARE INC

AI summary

A method for providing multi-tenancy support for RDMA in a system that includes a plurality of physical hosts. Each each physical host hosts a set of data compute nodes (DCNs). The method, at an RDMA protocol stack of the first host, receives a packet that includes a request from a first DCN hosted on a first host for RDMA data transfer from a second DCN hosted on a second host. The method sends a set of parameters of an overlay network that are associated with the first DCN to an RDMA physical network interface controller of the first host. The set of parameters are used by the RDMA physical NIC to encapsulate the packet with an RDMA data transfer header and an overlay network header by using the set of parameters of the overlay network to transfer the encapsulated packet to the second physical host using the overlay network.