Miscellaneous Changes/Fixes (#325)

QUICK_START.md

@@ -200,9 +200,9 @@ poetry run caribou deploy_remote_cli
 ```
 
 You may also specify the `memory` (in MB), `timeout` (in seconds), and `ephemeral_storage` (in MB) using the following flags:
- - `memory`: Use `--memory` or `-m`. Default: 5,120 MB
+ - `memory`: Use `--memory` or `-m`. Default: 1,769 MB (1 Full vCPU)
  - `timeout`: Use `--timeout` or `-t`. Default: 900 seconds
- - `ephemeral_storage`: Use `--ephemeral_storage` or `-s`. Default: 10,240 MB
+ - `ephemeral_storage`: Use `--ephemeral_storage` or `-s`. Default: 5,120 MB
 
 To remove the remote framework, use the following command:
 

caribou-go/src/deployment-metrics-calculator/deployment-metrics-calculator.go

@@ -102,7 +102,7 @@ type SimpleDeploymentMetricsCalculator struct {
 func SetupSimpleDeploymentMetricsCalculator(dataString string) *SimpleDeploymentMetricsCalculator {
 	return &SimpleDeploymentMetricsCalculator{
 		*SetupDeploymentMetricsCalculator(dataString),
-		1000,
+		200,
 	}
 }
 
@@ -144,7 +144,7 @@ func (sd *SimpleDeploymentMetricsCalculator) PerformMonteCarloSimulation(deploym
 	executionCarbonList := []float64{}
 	transmissionCarbonList := []float64{}
 
-	maxNumberOfIterations := 20000
+	maxNumberOfIterations := 2000
 	threshold := 0.05
 	numberOfIterations := 0
 

caribou-go/src/deployment-metrics-calculator/deployment-metrics-calculator_test.go

@@ -33,5 +33,5 @@ func TestSimpleDeploymentMetricsCalculator_PerformMonteCarloSimulation(t *testin
 	assert.Equal(t, result["tail_cost"], 1.0)
 	assert.Equal(t, result["tail_runtime"], 1.0)
 	assert.Equal(t, result["tail_carbon"], 1.0)
-	assert.Equal(t, callCounter, 20000)
+	assert.Equal(t, callCounter, 2000)
 }

caribou/deployment/client/cli/cli.py

@@ -156,7 +156,7 @@ def deploy_remote_cli(
         )
 
     ## Memory
-    memory_mb: int = 5120  # 5 GB (Not the maximum)
+    memory_mb: int = 1769  # 1 full vCPU (https://docs.aws.amazon.com/lambda/latest/dg/configuration-memory.html)
     if memory is not None:
         # Convert to int
         memory_mb = int(memory)
@@ -178,7 +178,7 @@ def deploy_remote_cli(
             raise click.ClickException("Timeout must be between 1 second and 900 seconds (15 minutes).")
 
     ## Ephemeral Storage
-    ephemeral_storage_mb = 10240  # Maximum ephemeral storage (10 GB)
+    ephemeral_storage_mb = 5120  # 5 GB (Should be enough for most use cases)
     if ephemeral_storage is not None:
         # Convert to int
         ephemeral_storage_mb = int(ephemeral_storage)

caribou/deployment_solver/deployment_metrics_calculator/simple_deployment_metrics_calculator.py

@@ -75,7 +75,7 @@ def __init__(
             record_transmission_execution_carbon,
         )
         self.n_processes = n_processes
-        self.batch_size = 1000
+        self.batch_size = 200
         if n_processes > 1:
             self._setup(
                 workflow_config,
@@ -190,7 +190,7 @@ def _perform_monte_carlo_simulation(self, deployment: list[int]) -> dict[str, fl
         execution_carbon_list: list[float] = []
         transmission_carbon_list: list[float] = []
 
-        max_number_of_iterations = 20000
+        max_number_of_iterations = 2000
         threshold = 0.05
         number_of_iterations = 0
         while number_of_iterations < max_number_of_iterations:

caribou/syncers/components/execution_data.py

@@ -63,11 +63,16 @@ def _get_total_output_data_size(self) -> float:
 
         return total_output_data_size
 
-    def _get_total_input_data_size(self) -> float:
+    def _get_total_input_data_size(self, consider_input_payload: bool = False) -> float:
         total_input_data_size = 0.0
 
-        if self.input_payload_size:
-            total_input_data_size += self.input_payload_size
+        # It appears that the input payload size is not captured
+        # in the lambda insights, so we will not include it in the
+        # total input data size calculation, if it is not explicitly
+        # requested.
+        if consider_input_payload:
+            if self.input_payload_size:
+                total_input_data_size += self.input_payload_size
 
         if self.download_size:
             total_input_data_size += self.download_size

caribou/tests/deployment_solver/deployment_metrics_calculator/test_simple_deployment_metrics_calculator.py

@@ -84,7 +84,7 @@ def test_perform_monte_carlo_simulation(self, mock_calculate_workflow):
         self.assertEqual(results["tail_carbon"], 1.0)
 
         # Verify that the mock method was called the correct number of times with the correct arguments
-        self.assertEqual(mock_calculate_workflow.call_count, 20000)
+        self.assertEqual(mock_calculate_workflow.call_count, 2000)
         mock_calculate_workflow.assert_called_with(deployment)
 
     @patch.object(

caribou/tests/syncers/components/test_execution_data.py

@@ -37,12 +37,13 @@ def test_get_total_output_data_size(self):
         self.assertEqual(self.execution_data._get_total_output_data_size(), 10.0)
 
     def test_get_total_input_data_size(self):
+        # Not part of total input data size unless explicitly specified
         self.execution_data.input_payload_size = 1.0
         self.execution_data.download_size = 2.0
         successor_data = Mock(spec=ExecutionToSuccessorData)
         successor_data.get_total_input_data_size.return_value = 3.0
         self.execution_data.successor_data["successor1"] = successor_data
-        self.assertEqual(self.execution_data._get_total_input_data_size(), 6.0)
+        self.assertEqual(self.execution_data._get_total_input_data_size(), 5.0)
 
     def test_data_transfer_during_execution(self):
         self.execution_data.lambda_insights = {"total_network": 10 * 1024**3}  # 10 GB
@@ -52,7 +53,7 @@ def test_data_transfer_during_execution(self):
         successor_data.get_total_input_data_size.return_value = 3.0
         successor_data.get_total_output_data_size.return_value = 4.0
         self.execution_data.successor_data["successor1"] = successor_data
-        self.assertEqual(self.execution_data.data_transfer_during_execution, 10 - 6 - 4)
+        self.assertEqual(self.execution_data.data_transfer_during_execution, 10 - 5 - 4)
 
     def test_longest_duration(self):
         self.execution_data.user_execution_duration = 1.0
@@ -101,13 +102,13 @@ def test_to_dict(self):
             "duration_s": 3.0,
             "cpu_model": "test_cpu",
             "provider_region": "us-west-2",
-            "data_transfer_during_execution_gb": 10 - 6 - 4,
+            "data_transfer_during_execution_gb": 10 - 5 - 4,
             "cpu_utilization": 1.0,
             "successor_data": {"successor1": {"key": "value"}},
             "additional_analysis_data": {
                 "input_payload_size_gb": 1.0,
                 "download_size_gb": 2.0,
-                "total_input_data_transfer_gb": 6.0,
+                "total_input_data_transfer_gb": 5.0,
                 "total_output_data_transfer_gb": 4.0,
             },
         }