diff --git a/modin/core/execution/ray/implementations/pandas_on_ray/partitioning/virtual_partition.py b/modin/core/execution/ray/implementations/pandas_on_ray/partitioning/virtual_partition.py
index 65f2704d054..71058d97921 100644
--- a/modin/core/execution/ray/implementations/pandas_on_ray/partitioning/virtual_partition.py
+++ b/modin/core/execution/ray/implementations/pandas_on_ray/partitioning/virtual_partition.py
@@ -26,13 +26,6 @@
 from modin.utils import _inherit_docstrings
 
 
-# If Ray has not been initialized yet by Modin,
-# it will be initialized when calling `RayWrapper.put`.
-_DEPLOY_AXIS_FUNC = RayWrapper.put(PandasDataframeAxisPartition.deploy_axis_func)
-_DEPLOY_SPLIT_FUNC = RayWrapper.put(PandasDataframeAxisPartition.deploy_splitting_func)
-_DRAIN = RayWrapper.put(PandasDataframeAxisPartition.drain)
-
-
 class PandasOnRayDataframeVirtualPartition(PandasDataframeAxisPartition):
     """
     The class implements the interface in ``PandasDataframeAxisPartition``.
@@ -60,6 +53,32 @@ class PandasOnRayDataframeVirtualPartition(PandasDataframeAxisPartition):
     instance_type = ray.ObjectRef
     axis = None
 
+    _DEPLOY_AXIS_FUNC = None
+    _DEPLOY_SPLIT_FUNC = None
+    _DRAIN_FUNC = None
+
+    @classmethod
+    def _get_deploy_axis_func(cls):
+        if cls._DEPLOY_AXIS_FUNC is None:
+            cls._DEPLOY_AXIS_FUNC = RayWrapper.put(
+                PandasDataframeAxisPartition.deploy_axis_func
+            )
+        return cls._DEPLOY_AXIS_FUNC
+
+    @classmethod
+    def _get_deploy_split_func(cls):
+        if cls._DEPLOY_SPLIT_FUNC is None:
+            cls._DEPLOY_SPLIT_FUNC = RayWrapper.put(
+                PandasDataframeAxisPartition.deploy_splitting_func
+            )
+        return cls._DEPLOY_SPLIT_FUNC
+
+    @classmethod
+    def _get_drain_func(cls):
+        if cls._DRAIN_FUNC is None:
+            cls._DRAIN_FUNC = RayWrapper.put(PandasDataframeAxisPartition.drain)
+        return cls._DRAIN_FUNC
+
     def __init__(
         self,
         list_of_partitions,
@@ -169,7 +188,7 @@ def deploy_splitting_func(
             if extract_metadata
             else num_splits,
         ).remote(
-            _DEPLOY_SPLIT_FUNC,
+            cls._DEPLOY_SPLIT_FUNC,
             axis,
             func,
             f_args,
@@ -230,7 +249,7 @@ def deploy_axis_func(
             * (1 + cls._PARTITIONS_METADATA_LEN),
             **({"max_retries": max_retries} if max_retries is not None else {}),
         ).remote(
-            _DEPLOY_AXIS_FUNC,
+            cls._DEPLOY_AXIS_FUNC,
             axis,
             func,
             f_args,
@@ -486,7 +505,7 @@ def drain_call_queue(self, num_splits=None):
             _ = self.list_of_blocks
             return
         drained = super(PandasOnRayDataframeVirtualPartition, self).apply(
-            _DRAIN, num_splits=num_splits, call_queue=self.call_queue
+            self._DRAIN, num_splits=num_splits, call_queue=self.call_queue
         )
         self._list_of_block_partitions = drained
         self.call_queue = []