Read Request Scheduling for Tape Libraries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current scheduling techniques for read access in enterprise tape libraries are inefficient and slow, particularly for high latency storage media like tapes, due to slow average seek times and limitations to single storage media.
Innovation Solution
A method that organizes read requests based on deadlines and selects replicas for objects on storage media, generating an initial schedule that minimizes latency and optimizes execution time by considering availability and constraints like load/unload delays and seek latency, using a data management computing device to manage read access across multiple drives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If current scheduling techniques are used for read access in enterprise tape libraries, then the system operates with existing scheduling methods, but the scheduling efficiency is low and execution time is slow due to slow average seek times
Solution Approach 1:
The system performs preliminary actions by pre-loading tapes onto drives before they are actually needed for read operations. The scheduling algorithm anticipates future read requests and proactively positions media on drives, thereby reducing the average seek time and improving scheduling efficiency when actual read operations occur.
Solution Approach 2:
The scheduling algorithm dynamically adjusts the scheduling of read requests based on current system state, including drive availability, media location, and request priorities. This dynamic scheduling optimizes the sequence of operations to minimize seek times and improve overall productivity while reducing execution time.
2Adaptability or versatility
If current scheduling techniques are used, then the system maintains simplicity in scheduling logic, but the techniques are limited to single storage media and cannot optimize across multiple media
Solution Approach 1:
The scheduling algorithm is designed to work across multiple types of storage media (tapes, disks, etc.) and can select from multiple replicas of the same object stored on different media. This universal approach allows the system to optimize read operations by choosing the best available replica regardless of media type, enhancing adaptability without requiring separate scheduling logic for each media type.
Solution Approach 2:
The system utilizes multiple replicas of objects stored on different media. When a read request arrives, the scheduler can select from multiple copies, choosing the one that will provide the fastest access based on current media availability and drive status. This copying strategy enables multi-media optimization while keeping the scheduling logic manageable.
3Productivity
If more load/unload operations are performed on storage media, then more read requests can be serviced, but the reliability decreases due to increased wear and potential failures
Solution Approach 1:
The system performs preliminary loading of tapes onto drives in advance, so that when read requests arrive, the media is already in position and ready for immediate access. This reduces the need for frequent load/unload operations during normal operation, thereby maintaining reliability while still achieving high throughput for read requests.
Solution Approach 2:
The scheduler maintains continuous utilization of loaded media by planning and executing read operations in sequences that keep tapes on drives for extended periods. This continuity maximizes the useful action extracted from each loaded tape, increasing throughput without proportionally increasing load/unload operations, thus preserving reliability.
Data Source
AI summary
A method, device and non-transitory computer readable medium that manages read access includes organizing a plurality of requests for objects on one or more storage media, such as tapes or spin-down disks, based on at least a deadline for each of the plurality of requests. One of one or more replicas for each of the objects on the one or more storage media is selected based on one or more factors. An initial schedule for read access is generated based at least on the deadline for each of the plurality of requests, the selected one of the replicas for each of the objects, and availability of one or more drives. The initial schedule for read access on the one or more of the drives for each of the plurality of requests for the objects is provided.