STOW RS Upload performance measurements Parallel TCP Connections - dcm4che/dcm4chee-arc-light GitHub Wiki
Enable to send several hundreds or thousands of DICOM files or huge datasets of DICOM data to STOW-RS server with(-out) splitting them into several http requests sent over several TCP connections.
Upload performance measurements have been done for concurrently invoked STOW RS services wherein each request contains either 5 files or 500MB data for different types of modality objects. No limit has been used which implies that each TCP connection contains only one HTTP multipart request in it.
| Parallel TCP Connections | Description |
|---|---|
| 2 | Store 10 objects or 1000MB data using 2 parallel connections |
| 4 | Store 20 objects or 2000MB data using 4 parallel connections |
| 6 | Store 30 objects or 3000MB data using 6 parallel connections |
| 8 | Store 40 objects or 4000MB data using 8 parallel connections |
Modality objects data used in tests
| Modality Type | Size per object |
|---|---|
| MR | 513.3KB |
| CT | 1.9MB |
| CR | 5.9MB |
| DX | 11MB |
Start stowrsd tool acting as STOW-RS server
stowrsd -b 8080 --ignore
- Parallel STOW-RS invocations with each request containing 5 files
-
No. of Parallel TCP Connections vs Time in millisec
-
No. of Parallel TCP Connections vs MB/s
- Parallel STOW-RS invocations with each request containing 500MB
-
No. of Parallel TCP Connections vs Time in sec
-
No. of Parallel TCP Connections vs MB/s
