Sliding Window Switch Scheduling for Fast Crossbar Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing switching algorithms for input-queued crossbar switches face challenges in achieving high-quality matchings within a short time frame, leading to suboptimal switch throughput and queueing delays, particularly in high-speed networks like 256-port switches with 100 Gbps line rates.
Innovation Solution
A sliding window scheduling method and system, such as SW-QPS and SW-iSLIP, that utilize bitmap operators and distributed parallelization hardware to achieve O(1) time complexity per matching computation, enabling efficient packet scheduling through queue-proportional sampling and first-fit-accepting policies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional switching algorithms are used to compute matchings, then high-quality matchings can be achieved, but the time complexity is too high to complete computation within a short time slot
Solution Approach 1:
The patent segments the matching computation into multiple iterations, where each iteration computes a partial matching. The iSLIP algorithm divides the matching process into request, grant, and accept phases, allowing incremental computation of matchings that can be completed within tight time slots while maintaining high quality through multiple refinement iterations.
Solution Approach 2:
The patent uses preliminary actions by pre-computing availability information and queue lengths before the actual matching computation. Input ports prepare request messages with pre-calculated availability bitmaps, and output ports pre-process grant messages, allowing the core matching algorithm to operate on prepared data structures rather than raw queue states.
2Productivity
If switching algorithms compute high-quality matchings, then switch throughput is improved, but queueing delays increase due to computation time
Solution Approach 1:
The patent implements periodic action through time-division multiplexing of switching phases. Within each scheduling cycle, the system alternates between computation phases (request, grant, accept) and switching phases, allowing multiple matching computations to be performed periodically over time. This enables high throughput through accumulated matchings while keeping individual queueing delays bounded by the periodic cycle time.
Solution Approach 2:
The patent maintains continuity of useful action by ensuring that the crossbar switching operation continues uninterrupted during matching computations. While control logic performs matching computations, the crossbar simultaneously switches packets based on previously computed matchings, eliminating idle periods and maintaining continuous productive operation that improves throughput without increasing queueing delays.
Data Source
AI summary
An exemplary sliding window scheduling method and system are disclosed. The exemplary sliding window scheduling method and system can schedule multiple packets in a given scheduling frame with a sliding window scheduling frame. The scheduling operation can be performed using bitmap operators and can achieve a lowest time complexity of O(1) per matching computation and per port using distributed parallelization hardware. The exemplary sliding window scheduling method and system can be performed in the context of a queue-proportional scheduler (QPS) as well as iSLIP. In alternative embodiments, the SW-QPS operation can be performed in a batching window rather than in a sliding window.