Mixed DiT Serving on Skewed workload mix 12 req/min, SLO Scale = 1.0x FLUX on H100
75SLO Attainment Ratio (SAR)TetriServe
Evaluation Results
| Method | Links | |
|---|---|---|
| TetriServeNirvana Integration=true, Hardware=H100, Model=FLUX2025.10 | 75 | |
| RSSPNirvana Integration=true, Hardware=H100, Model=FLUX2025.10 | 53 | |
| TetriServeNirvana Integration=false, Hardware=H100, Model=FLUX2025.10 | 19 | |
| RSSPNirvana Integration=false, Hardware=H100, Model=FLUX2025.10 | 4 |