Performance Terms¶
Understanding Azure Storage performance metrics is critical for optimizing application throughput and latency.
Note
Compare server latency and end-to-end latency together to separate backend pressure from client/network delay.
Key Performance Metrics¶
| Term | Definition | Context |
|---|---|---|
| IOPS | Input/Output Operations Per Second | Small random reads/writes |
| Throughput | Data transfer rate (MB/s or GB/s) | Large sequential transfers |
| Latency | Time to complete a single request | Measured in milliseconds |
| E2E Latency | Total time from request start to finish | Includes network transit |
| Server Latency | Time spent processing on Azure side | Excludes network transit |
| Ingress | Data flowing into the storage account | Uploads |
| Egress | Data flowing out of the storage account | Downloads |
| Transaction | A single REST API call | Billing and limit unit |
| Partition | Logical grouping of data for scale | Scaling unit |
| Throttling | Rate limiting when targets are exceeded | HTTP 429 errors |
| Bursting | Temporary boost above provisioned limit | Premium file shares |
| Premium | SSD-based high-performance tier | Low latency |
| Standard | HDD-based general-purpose tier | Cost-effective |
| Hot/Cool Tier | Access tiers for varying frequency | Performance/Cost trade-off |
| Cold | Online access tier | Immediate access; milliseconds latency |
| Archive | Offline access tier | Rehydration required; hours latency |
Performance Concepts Relationship¶
```mermaid graph LR Perf[Performance] --> Lat[Latency] Perf --> TP[Throughput] Perf --> IOPS[IOPS]
Lat --> Srv[Server Latency]
Lat --> Net[Network Latency]
TP --> In[Ingress]
TP --> Out[Egress]
Lim[Limits] --> Thr[Throttling]
Lim --> Bst[Bursting]
```