Shared Reserve Memory Pool for Network Queue Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In network devices, low-activity queues can become ineligible for memory allocation due to high demand from busy queues, leading to congestion or packet drops, as existing allocation schemes prioritize busy queues and reduce the MAX-STATE threshold, leaving low-activity queues without sufficient resources to handle bursts.
Innovation Solution
A memory management circuit that maintains a shared-reserve memory pool, identifying queues that require additional resources but are ineligible for the shared buffer, and allocates from this pool if the occupancy is within a defined margin above the MAX-STATE threshold, ensuring fair resource distribution and preventing starvation of low-activity queues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing allocation schemes prioritize busy queues and reduce the MAX-STATE threshold, then memory resources are efficiently allocated to high-demand queues, but low-activity queues become ineligible for memory allocation and may suffer congestion or packet drops
Solution Approach 1:
The memory pool is segmented into two distinct parts: a shared buffer for dynamic allocation to queues based on demand, and a shared-reserve memory pool reserved specifically for low-activity queues. This segmentation allows busy queues to efficiently utilize the shared buffer while ensuring low-activity queues have guaranteed access to the reserve pool, preventing starvation and maintaining service reliability.
Solution Approach 2:
Different memory allocation policies are applied to different queue types based on their activity characteristics. High-activity queues receive dynamic allocation from the shared buffer with adaptive MAX-STATE thresholds, while low-activity queues are guaranteed access to the shared-reserve memory pool. This local quality approach ensures each queue type receives appropriate resources for its specific needs.
2Quantity of substance
If a shared buffer is dynamically allocated to queues based on demand, then memory utilization is optimized for busy queues, but low-activity queues may be starved of resources when the shared buffer is fully allocated
Solution Approach 1:
The shared-reserve memory pool is pre-established and reserved specifically for low-activity queues before any allocation decisions are made. This preliminary action ensures that when the shared buffer becomes fully allocated to busy queues, low-activity queues still have guaranteed access to reserve resources, preventing starvation and maintaining system adaptability.
Solution Approach 2:
The shared-reserve memory pool acts as an intermediary resource that mediates between the shared buffer and low-activity queues. When the shared buffer is exhausted, the reserve pool provides an intermediate allocation mechanism that ensures low-activity queues can still receive memory resources, bridging the gap between dynamic allocation efficiency and guaranteed resource availability.
3Productivity
If the MAX-STATE threshold is reduced to prioritize busy queues, then high-demand queues receive sufficient memory resources, but low-activity queues become ineligible for allocation even when buffer space is available
Solution Approach 1:
The memory allocation system is segmented into two independent pathways: the shared buffer with reduced MAX-STATE thresholds for busy queues, and the shared-reserve memory pool with separate eligibility criteria for low-activity queues. This segmentation allows aggressive threshold reduction for high-demand queues without preventing low-activity queues from accessing reserve resources, maintaining both throughput and accessibility.
Solution Approach 2:
Different MAX-STATE threshold parameters are applied to different queue types. Busy queues operate with reduced thresholds to prioritize throughput, while low-activity queues have separate eligibility parameters that allow access to the reserve pool. This parameter differentiation resolves the contradiction by allowing threshold reduction without eliminating allocation accessibility for all queues.
Data Source
AI summary
A device includes ports, a packet processor, and a memory management circuit. The ports communicate packets over a network. The packet processor processes the packets using queues. The memory management circuit maintains a shared buffer in a memory and adaptively allocates memory resources from the shared buffer to the queues, maintains in the memory, in addition to the shared buffer, a shared-reserve memory pool for use by the queues, identifies, among the queues, a queue that requires additional memory resources, the queue having an occupancy that is (i) above a current value of a dynamic threshold, rendering the queue ineligible for additional allocation from the shared buffer, and (ii) no more than a defined margin above the current value of the dynamic threshold, rendering the queue eligible for allocation from the shared-reserve memory pool, and allocates memory resources to the identified queue from the shared-reserve memory pool.