[2/N] Elastic EP Milestone 2: Integrating NIXL-EP (#35627)

Signed-off-by: Itay Alroy <ialroy@nvidia.com>
Co-authored-by: Yongji Wu <wuyongji317@gmail.com>
Co-authored-by: Ron Tourgeman <rtourgeman@nvidia.com>

vllm/envs.py

@@ -244,6 +244,7 @@ if TYPE_CHECKING:
     VLLM_ELASTIC_EP_SCALE_UP_LAUNCH: bool = False
     VLLM_ELASTIC_EP_DRAIN_REQUESTS: bool = False
     VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: bool = False
+    VLLM_NIXL_EP_MAX_NUM_RANKS: int = 32
 
 
 def get_default_cache_root():
@@ -1628,6 +1629,10 @@ environment_variables: dict[str, Callable[[], Any]] = {
     "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": lambda: bool(
         int(os.getenv("VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS", "0"))
     ),
+    # NIXL EP environment variables
+    "VLLM_NIXL_EP_MAX_NUM_RANKS": lambda: int(
+        os.getenv("VLLM_NIXL_EP_MAX_NUM_RANKS", "32")
+    ),
 }