Prediction-Based Switch Allocator Arbitration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Switch allocators in on-chip and inter-chip routers face challenges in maximizing router throughput while maintaining fairness among input and output ports, often resulting in arbitration collisions that limit network performance due to the inability to acquire a global snapshot of input requests within a clock cycle, leading to wasted bandwidth.

Innovation Solution

An arbitration mechanism that maintains mappings from preceding arbitration decision cycles, using a pipeline to selectively couple input and output ports based on previous requests, reducing the solution space and minimizing collisions through a prediction-based technique.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If distributed arbitration is used to avoid global snapshot requirements, then the switch allocator can operate at current clock speeds without acquiring global snapshots, but arbitration collisions occur leading to wasted bandwidth

Engineering Contradiction:
Improverouter throughputVSAvoidwasted bandwidth
Core Design Contradiction:
ProductivityVSLoss of energy

Solution Approach 1:

The patent applies preliminary action by maintaining mappings from preceding arbitration decision cycles and using them in prediction-based allocation. The switch allocator predicts future arbitration decisions based on historical mappings, allowing it to proactively assign output ports before conflicts occur, thereby avoiding wasted bandwidth while maintaining high throughput without requiring global snapshots

Inventive Principle:
Principle #10Preliminary action

2Productivity

If the switch allocator maximizes the number of flits transferred, then router throughput increases, but fairness among input and output ports deteriorates

Engineering Contradiction:
Improverouter throughputVSAvoidfairness among ports
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent applies dynamics by making the arbitration decision adaptive based on historical mappings and current requests. The prediction-based allocator dynamically adjusts its decisions by considering both the maximization of flit transfers and the maintenance of fairness, using feedback from previous cycles to balance throughput optimization with equitable resource distribution among ports

Inventive Principle:
Principle #15Dynamics

3Reliability

If input ports are restricted to transferring at most one flit per arbitration decision cycle, then the switch allocator can maintain fairness constraints, but the number of flits transferred per cycle is limited

Engineering Contradiction:
Improvefairness constraintsVSAvoidflits transferred per cycle
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent applies preliminary action by pre-establishing mappings from previous arbitration cycles that respect the one-flit constraint. By maintaining historical mappings and using prediction-based allocation, the system can maximize flit transfers within the fairness constraint by strategically assigning ports based on predicted future needs, thereby achieving high productivity without violating the per-port flit limit

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS8976802B2Prediction-based switch allocator
Publication Date: 2015.03.10 ORACLE INT CORP
  • US8976802B2 patent drawing
  • US8976802B2 patent drawing
  • US8976802B2 patent drawing

AI summary

An arbitration technique for determining mappings for a switch is described. During a given arbitration decision cycle, an arbitration mechanism maintains, until expiration, a set of mappings from a subset of the input ports to a subset of the output ports of the switch. This set of mappings was determined during an arbitration decision cycle up to K cycles preceding the given arbitration decision cycle. Because the set of mappings are maintained, it is easier for the arbitration mechanism to determine mappings from a remainder of the input ports to the remainder of the output ports without collisions.