[rhaiis] Add named workload aliases, prefix-cache toggles, and composite benchmark presets

projects/rhaiis/orchestration/config.d/workloads.yaml

@@ -40,6 +40,48 @@ profile7:
   rates: [1, 50, 100, 200, 300]
   max_seconds: 450
 
+# ── Named aliases ───────────────────────────────────────────
+# Meaningful names that map to the numbered profiles above.
+# Use these in presets for readability; the underlying config
+# is identical to the corresponding profileN.
+
+balanced:
+  data: "prompt_tokens=1000,output_tokens=1000"
+  rates: [1,50,100,200,300]
+  max_seconds: 450
+
+variable:
+  data: "prompt_tokens=512,prompt_tokens_stdev=128,prompt_tokens_min=1,prompt_tokens_max=1024,output_tokens=2048,output_tokens_stdev=512,output_tokens_min=1,output_tokens_max=4096"
+  rates: [1,50,100,200,300]
+  max_seconds: 450
+
+summarization:
+  data: "prompt_tokens=2048,output_tokens=128"
+  rates: [1, 50, 100, 200, 300]
+  max_seconds: 450
+
+long-context:
+  data: "prompt_tokens=8000,output_tokens=1000"
+  rates: [1,25,50,75,100]
+  max_seconds: 450
+  samples: 50
+
+ultra-long-context:
+  data: "prompt_tokens=100000,output_tokens=1000"
+  rates: [1, 2, 5]
+  max_seconds: 450
+  samples: 10
+
+multi-turn:
+  data: "prompt_tokens=1000,output_tokens=1000,turns=5,prefix_tokens=512,prefix_count=10"
+  rates: [1, 25, 50, 75, 100]
+  max_seconds: 450
+
+heavy-heterogeneous:
+  data: "prompt_tokens=8000,prompt_tokens_stdev=8500,prompt_tokens_min=50,prompt_tokens_max=30000,output_tokens=800,output_tokens_stdev=1500,output_tokens_min=20,output_tokens_max=8000"
+  rates: [1, 50, 100, 200, 300]
+  max_seconds: 450
+
 # Dashboard override stub (set data, rates, max_seconds at runtime)
 custom:
   data: "prompt_tokens=1000,output_tokens=1000"

projects/rhaiis/orchestration/presets.d/benchmarks.yaml

@@ -6,3 +6,49 @@ benchmark:
   rhaiis.profiler.enabled: true
   caliper.postprocess.csv_dashboard.enabled: true
   rhaiis.agent_analysis.enabled: false
+
+benchmark-standard:
+  extends:
+  - benchmark
+  tests.rhaiis.workload_key:
+  - balanced
+  - variable
+
+benchmark-long-context:
+  extends:
+  - benchmark
+  tests.rhaiis.workload_key:
+  - long-context
+  - ultra-long-context
+
+benchmark-full-sweep:
+  extends:
+  - benchmark
+  tests.rhaiis.workload_key:
+  - balanced
+  - variable
+  - summarization
+  - long-context
+  - ultra-long-context
+  - multi-turn
+  - heavy-heterogeneous
+
+benchmark-multi-turn:
+  extends:
+  - benchmark
+  tests.rhaiis.workload_key: multi-turn
+
+benchmark-multi-turn-prefix-on:
+  extends:
+  - benchmark-multi-turn
+  - prefix-cache-on
+
+benchmark-multi-turn-prefix-off:
+  extends:
+  - benchmark-multi-turn
+  - prefix-cache-off
+
+benchmark-heterogeneous:
+  extends:
+  - benchmark
+  tests.rhaiis.workload_key: heavy-heterogeneous

projects/rhaiis/orchestration/presets.d/presets.yaml

@@ -63,6 +63,37 @@ profile6:
 profile7:
   tests.rhaiis.workload_key: profile7
 
+# Named aliases (same workloads, readable names)
+balanced:
+  tests.rhaiis.workload_key: balanced
+
+variable:
+  tests.rhaiis.workload_key: variable
+
+summarization:
+  tests.rhaiis.workload_key: summarization
+
+long-context:
+  tests.rhaiis.workload_key: long-context
+
+ultra-long-context:
+  tests.rhaiis.workload_key: ultra-long-context
+
+multi-turn:
+  tests.rhaiis.workload_key: multi-turn
+
+heavy-heterogeneous:
+  tests.rhaiis.workload_key: heavy-heterogeneous
+
+# ── Feature toggles ─────────────────────────────────────────
+
+prefix-cache-on:
+  rhaiis.engines.vllm.args.no-enable-prefix-caching: false
+  rhaiis.engines.vllm.args.enable-prefix-caching: true
+
+prefix-cache-off:
+  rhaiis.engines.vllm.args.no-enable-prefix-caching: true
+
 # ── Models ──────────────────────────────────────────────────
 # Short aliases default to the FP8 variant where available.
 # Use the full name (e.g. llama-70b-w4a16) for other quants.