Statistical Admission Control for Disk Bandwidth Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Magnetic disk drives face inefficiencies in managing multiple video and audio streams due to complexity in cycle-based scheduling and limited bandwidth, particularly when handling variable bit rate streams and concurrent reading/writing operations, which can lead to missed deadlines and resource constraints.
Innovation Solution
The implementation of a statistical-based admission control technique, TRAC, which models and dynamically adjusts bandwidth allocation using three random variables to optimize disk bandwidth sharing, supporting variable bit rate streams and concurrent operations by defining a maximum probability of missed deadlines and efficiently partitioning disk bandwidth between reading and writing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If cycle based scheduling is used to guarantee service quality, then reliability is improved, but device complexity increases due to worst-case parameter assumptions
Solution Approach 1:
The patent transitions from deterministic worst-case parameter assumptions to statistical parameter distributions. By modeling disk parameters (transfer rate, seek time, rotational latency) as random variables with known distributions, the system achieves reliability through statistical guarantees rather than conservative worst-case bounds, reducing complexity while maintaining service quality.
Solution Approach 2:
The patent replaces the mechanical cycle-based scheduling approach with a statistical admission control mechanism. Instead of rigid time cycles and worst-case timing analysis, the system uses probabilistic models and statistical thresholds to manage stream admissions, simplifying the scheduling logic while preserving reliability through statistical service level agreements.
2Productivity
If random placement with deadline driven approach is used, then productivity is improved, but reliability worsens due to missed deadlines in multi-stream scenarios
Solution Approach 1:
The patent implements feedback through statistical admission control that monitors system state and adjusts admission decisions accordingly. By calculating the probability of meeting deadlines based on current stream admissions and disk parameter distributions, the system provides feedback-driven admission decisions that maintain high productivity while ensuring reliability through probabilistic deadline guarantees.
Solution Approach 2:
The patent makes the admission control dynamic by using real-time statistical analysis of disk parameter distributions and current system load. Instead of static admission thresholds, the system dynamically adjusts admission decisions based on the statistical likelihood of meeting deadlines under current conditions, balancing productivity and reliability adaptively.
3Reliability
If admission control limits streams to minimize missed deadlines, then reliability is improved, but productivity decreases due to restricted bandwidth utilization
Solution Approach 1:
The patent transforms the admission control from a conservative stream-limiting approach to a probabilistic bandwidth allocation approach. By modeling bandwidth consumption and disk parameters as random variables, the system determines admission decisions based on statistical bandwidth utilization thresholds rather than fixed stream limits, maximizing productivity while maintaining reliability through statistical guarantees.
Data Source
AI summary
Admission of a new disk stream is based on the probability of overcommitting disk bandwidth based on parameters related to the disk. These fixed parameters are determined either by retrieval from the disk or by investigating the disk. Probability functions of the disk parameters may be obtained. Exemplary disk parameters may be average disk seek time, probabilistic determination of the amount of data exchange during a single exchange, and probabilistic information about reading versus writing.