Table of Contents
The UC Berkeley ISTORE Project: bringing availability, maintainability, and evolutionary growth to storage-based clusters
Outline
The problem space: big data
One approach: traditional NAS
The real scalability problems: AME
The ISTORE project vision
Outline
Principles for achieving AME (1)
Principles for achieving AME (2)
Outline
Hardware techniques
Hardware techniques (2)
Hardware techniques (3)
ISTORE-1 hardware platform
A glimpse into the future?
Software techniques
Software techniques (2)
Software techniques (3)
Software techniques (4)
Applications
Outline
Availability benchmarks
Availability benchmark methodology
Methodology: reporting results
Example results: software RAID-5
Example results: multiple-faults
Benchmark conclusions
ISTORE conclusions
Future work
PPT Slide
Backup Slides
Case study
Benchmark environment: metrics
Benchmark environment: faults
Disk fault injection technique
System configuration
Results: single-fault experiments
System behavior: single-fault
System behavior: single-fault (2)
Interpretation: single-fault exp’ts
Results: multiple-fault experiments
Interpretation: multi-fault exp’ts
Maintainability Observations
Maintainability Observations (2)
|