Per-Priority MTU Sizing for Data Center Ethernet
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data center Ethernet networks face challenges in optimizing Maximum Transmission Unit (MTU) sizes, as increasing MTU uniformly for all queues requires larger memory reservations, leading to increased latency for certain types of network traffic, and existing solutions do not allow for efficient configuration based on varying traffic types.
Innovation Solution
Implementing per-priority MTU sizing within data center Ethernet networks using the Data Center Bridging Capabilities Exchange (DCBX) Link Layer Discovery Protocol, which allows for configurable MTU sizes for each priority level, optimizing memory allocation and reducing latency by tailoring MTU sizes to specific traffic types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If MTU size is increased uniformly for all queues, then network transmission efficiency is improved, but memory reservation requirements increase and latency increases for certain traffic types
Solution Approach 1:
The patent implements per-queue MTU sizing where each queue can have a different MTU configuration optimized for its specific traffic type. This allows high-priority latency-sensitive traffic to use smaller MTUs for faster processing while bulk data traffic uses larger MTUs for improved transmission efficiency, resolving the contradiction between overall efficiency and specific latency requirements.
Solution Approach 2:
The network traffic is segmented into multiple priority queues, each with independently configurable MTU sizes. This segmentation allows different MTU optimizations for different traffic classes, preventing the uniform MTU increase from causing latency issues across all traffic types while still improving overall network efficiency.
2Productivity
If MTU size is increased uniformly for all queues, then network transmission efficiency is improved, but memory allocation for buffering increases
Solution Approach 1:
Each queue is configured with a specific MTU size appropriate for its traffic type, allowing memory buffers to be allocated based on actual per-queue requirements rather than uniformly across all queues. This reduces total memory allocation while maintaining the transmission efficiency benefits for queues that need larger MTUs.
Solution Approach 2:
By segmenting memory allocation per queue rather than providing a single large uniform buffer, the system allocates memory resources more efficiently. Only queues requiring larger MTUs allocate corresponding memory, reducing total memory consumption while maintaining high transmission efficiency where needed.
3Productivity
If per-priority MTU sizing is implemented, then memory utilization is optimized and latency is reduced, but configuration complexity increases
Solution Approach 1:
The system automatically discovers and configures appropriate MTU sizes for each queue based on traffic characteristics and requirements, reducing manual configuration complexity. The intelligent configuration mechanisms automatically optimize per-queue MTU settings without requiring complex manual setup, maintaining high network performance while simplifying administration.
Data Source
AI summary
A method of configuring a port on a network device includes receiving a frame of information, determining that the frame includes maximum transmission unit (MTU) information, and configuring a prioritized receive queue on the port with an MTU size based on the MTU information. The MTU size is different than another MTU size of another prioritized receive queue on the port. A network interface includes a port, a de-multiplexer coupled to an output or the port, and prioritized receive queues that each have an MTTU size. A first prioritized receive queue has an MTU size different than a second prioritized receive queue.