Layer-2 Switch Extended Addressing for InfiniBand Subnets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Infiniband networks face limitations in address space, particularly in High-Performance Computing (HPC) networks and data centers, where the 16-bit Destination Local Identifier (DLID) restricts the number of addressable entities per subnet to 48K, necessitating additional network elements and administrative overhead for larger address requirements.
Innovation Solution
The method involves assigning extended layer-2 addresses with n=m+k bits, where k>0, by combining layer-2 and layer-3 destination addresses to reconstruct the extended layer-2 address, allowing for increased addressability without requiring routers or changes to Host Channel Adapters (HCAs), enabling support for larger subnets while preserving existing subnet management practices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If the standard 16-bit DLID is used for layer-2 addressing, then the network structure remains simple and legacy systems are compatible, but the address space is limited to 48K entities per subnet
Solution Approach 1:
The patent merges layer-2 DLID (16 bits) with layer-3 DGID bits (k bits) to create an extended layer-2 address of n = 16 + k bits. This combining allows the network to utilize the existing DGID field in layer-3 packets to extend the addressing capability of layer-2 without requiring separate addressing mechanisms or additional network infrastructure.
Solution Approach 2:
The extended addressing mechanism serves multiple functions: it maintains compatibility with legacy 16-bit DLID systems while simultaneously supporting extended address spaces for larger subnets. The same packet structure and switching mechanism handle both traditional and extended addressing modes, making the system universally applicable to different network scales.
2Quantity of substance
If extended layer-2 addresses with n=16+k bits are implemented, then the address space expands beyond 48K limit, but the switching mechanism complexity increases
Solution Approach 1:
The switch is pre-configured with forwarding tables that map extended layer-2 addresses (combining DLID and DGID portions) to appropriate output ports. This preliminary configuration allows the switch to perform extended address lookup and forwarding without real-time computation or complex runtime processing, reducing operational complexity.
Solution Approach 2:
The patent introduces an intermediary mechanism where the layer-3 DGID field acts as a mediator to extend layer-2 addressing. Instead of creating a completely new layer-2 addressing system, the existing DGID structure is leveraged to provide extended addressing capability, simplifying the transition and reducing switching complexity.
3Quantity of substance
If additional network elements are introduced to expand address space, then addressability increases, but network complexity and administrative overhead increase
Solution Approach 1:
Instead of adding more network elements (horizontal expansion), the patent extends the address space by utilizing additional bits from the layer-3 DGID field (vertical/dimensional expansion). This approach expands addressing capability within the existing network structure without requiring additional routers, bridges, or other network infrastructure elements.
Data Source
AI summary
A communication method in a network operating in accordance with a standard that allocates a given number of bits m for layer-2 addressing of nodes in the network. The method includes accepting at a layer-2 switch in the network an assignment to one or more nodes in the network of respective layer-2 extended addresses, each including n=m+k bits, k>0. A given data packet is received at the switch for forwarding. The given data packet includes a layer-2 destination address and a layer-3 destination address in accordance with the standard. The layer-3 destination address includes t bits, t≧k. The given data packet is forwarded from the switch to one of the nodes by reading from the given data packet and combining the layer-2 destination address and k bits from the layer-3 destination address so as to reconstruct the n bits of the extended layer-2 address of the one of the nodes.