Network Traffic Management Device Interrupt Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network interface cards (NICs) face challenges in managing communications involving data transfer protocols with different throughput and latency characteristics, as they need to balance interrupt frequency to optimize performance for both high throughput and low latency protocols, which existing solutions do not effectively address.
Innovation Solution
An application delivery controller device with a network interface controller (NIC) that classifies data packets into high throughput and low latency queues, managing interrupts to optimize processing by storing packets in memory and determining when to interrupt processors based on protocol characteristics, allowing for efficient handling of both high throughput and low latency data transfers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If processor interrupts are frequent to handle low latency protocols, then latency is reduced, but throughput decreases due to excessive context switching
Solution Approach 1:
The patent segments network traffic into two distinct queues: a low latency queue for protocols requiring immediate processing (e.g., NFS) and a high throughput queue for protocols tolerant of batching (e.g., FTP). This segmentation allows the system to apply different interrupt handling strategies to each queue type, resolving the contradiction by treating latency-sensitive and throughput-sensitive traffic differently rather than using a single interrupt mechanism for all traffic.
Solution Approach 2:
The system dynamically adjusts interrupt handling behavior based on traffic characteristics. For the low latency queue, interrupts are generated immediately upon packet arrival to minimize delay. For the high throughput queue, interrupts are coalesced after accumulating multiple packets. This dynamic adaptation of interrupt frequency based on protocol requirements allows the system to optimize both latency and throughput performance simultaneously.
2Productivity
If processor interrupts are coalesced for high throughput protocols, then throughput is improved, but latency increases due to delayed processing
Solution Approach 1:
By dividing network traffic into separate low latency and high throughput queues, the patent enables targeted interrupt handling. The high throughput queue accepts packet coalescing and delayed interrupt generation to maximize throughput, while the low latency queue maintains immediate interrupt response. This segmentation resolves the contradiction by allowing throughput optimization without compromising latency-critical traffic.
Solution Approach 2:
Different quality attributes are applied to different traffic streams: the high throughput queue is optimized for maximum data transfer rate with tolerable latency, while the low latency queue is optimized for minimum delay with lower throughput requirements. This local quality differentiation allows each queue to be tuned independently for its specific performance goals.
3Device complexity
If a single interrupt mechanism is used for all protocols, then device complexity is reduced, but performance optimization for different protocol types is lost
Solution Approach 1:
The patent introduces a segmented queue structure with separate handling paths for low latency and high throughput traffic. While this adds some structural complexity, it enables significant performance optimization for different protocol types. The segmentation is implemented at the queue level rather than requiring complex per-protocol processing logic, maintaining relative simplicity while achieving protocol-specific optimization.
Solution Approach 2:
The dual-queue mechanism serves multiple functions: it acts as a simple pass-through for high throughput traffic while providing priority handling for low latency traffic. This multi-functionality allows a single interrupt handling framework to support both performance optimization modes without requiring entirely separate processing paths.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Handling network data packets classified as being high throughput and low latency with a network traffic management device is disclosed. Packets are received from a network and classified as high throughput or low latency based on packet characteristics or other factors. Low latency classified packets are generally processed immediately, such as upon receipt, while the low latency packet processing is strategically interrupted to enable processing coalesced high throughput classified packets in an optimized manner. The determination to cease processing low latency packets in favor of high throughput packets may be based on a number of factors, including whether a threshold number of high throughput classified packets are received or based on periodically polling a high throughput packet memory storage location.