Testing the Performance of Distributed Storage Systems
summary
Covers testing performance of Software Defined Storage (SDS) systems: explains SDS architectures, cache layers, why sync is expensive, common bottlenecks and which factors actually affect throughput and latency, and gives a detailed testing methodology. Also addresses data consistency and reliability trade-offs, warning that some SDS implementations sacrifice durability for raw performance.
AI-written summary · about
chapters (24)
- 0:00Introduction
- 1:58Test storage systems
- 4:31Test parameters
- 6:21Load scenarios and data
- 8:10RAID performance and working set
- 10:00Statistics for fair comparisons
- 11:27Sync operations importance
- 13:35Sync scalability
- 16:34Testing tools
- 17:15Cloud Storage testing
- 20:39Consistency in distributed systems
- 22:52High Flash Check
- 24:42Performance impact of consistency
- 26:53SSD reliability
- 27:58Mean time between failures
- 28:43Recovery time after failure
- 31:01Cluster reliability and recovery
- 33:40Cost vs scalability trade-offs
- 36:51File size limits
- 37:51Node failure recovery
- 40:47Compare with other systems
- 41:33File vs block storage
- 43:44Production deployment scale
- 44:26Quorum and network partitions
related talks
38mLoad Testing Storage Systems and the Peculiarities of Generating Test Data
40mBenchmarking: Repeatable & Comparable
40mPerformance Testing and Big Data
01h 08mPerformance testing in the cloud
44mBenchmark testing
40mTesting Distributed Systems w/ Deterministic Simulation
45mPerformance testing from scratch
27mUnderstanding Distributed Systems with Deterministic Simulation