Update README.md for more clear instructions and fix some minor errors.

Readme.md

@@ -42,6 +42,27 @@ For an example users should look at `las_scheduler.py` which implements Least At
 
 ### Running Blox
 
+#### Installation
+Blox uses gRpc, Matplotlib to communicate and Plot several collected Metric.
+We suggest the users to create a virtual environment to install the dependencies.
+```
+pip install grpcio
+pip install matplotlib
+pip install pandas==1.3.0
+pip install grpcio-tools
+
+pip install protobuf==4.21.1
+cd blox/deployment
+mkdir grpc_stubs
+make
+```
+
+#### Prepare the trace
+
+Take philly trace as an example, download the trace from [here](https://github.com/msr-fiddle/philly-traces/blob/master/trace-data.tar.gz) and unpack it.
+
+You can get a file named `job_cluster_log` which will be used in the following examples.
+
 Blox has two modes for running. One real cluster workload and second simulator. 
 
 ##### Simulation Mode
@@ -87,15 +108,6 @@ PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python python node_manager.py --ipaddr ip
 
 ### Details for reproducing results for artifacts
 These are instructions for reproducing artifacts for Blox.
-#### Installation 
-Blox uses gRpc, Matplotlib to communicate and Plot several collected Metric. 
-We suggest the users to create a virtual environment to install the dependencies.
-```
-pip install grpcio
-pip install matplotlib
-pip install pandas==1.3.0
-pip install grpcio-tools
-```
 
 ###### Running Blox Code
 To perform simulation.

placement/bebop.py

@@ -136,7 +136,7 @@ def _consolidated_placement(
                 # found a node with more GPUs then needed
                 if min_more_GPUs > len(free_gpus[node]):
                     min_more_GPUs = len(free_gpus[node])
-                    node_with_min_moRE_gpUs = node
+                    node_with_min_more_GPUs = node
         if node_with_min_more_GPUs is not None:
             # only extracting the GPUs we need
             return (free_gpus[node_with_min_more_GPUs][:numGPUs_needed], True)

placement/consolidated.py

@@ -108,7 +108,7 @@ def _consolidated_placement(
                 # found a node with more GPUs then needed
                 if min_more_GPUs > len(free_gpus[node]):
                     min_more_GPUs = len(free_gpus[node])
-                    node_with_min_moRE_gpUs = node
+                    node_with_min_more_GPUs = node
         if node_with_min_more_GPUs is not None:
             # only extracting the GPUs we need
             return (free_gpus[node_with_min_more_GPUs][:numGPUs_needed], True)

placement/consolidated_placement.py

@@ -106,7 +106,7 @@ def _consolidated_placement(
                 # found a node with more GPUs then needed
                 if min_more_GPUs > len(free_gpus[node]):
                     min_more_GPUs = len(free_gpus[node])
-                    node_with_min_moRE_gpUs = node
+                    node_with_min_more_GPUs = node
         if node_with_min_more_GPUs is not None:
             # only extracting the GPUs we need
             return (free_gpus[node_with_min_more_GPUs][:numGPUs_needed], True)

placement/first-gpu.py

@@ -249,7 +249,7 @@ def _consolidated_placement(
                 # found a node with more GPUs then needed
                 if min_more_GPUs > len(free_gpus[node]):
                     min_more_GPUs = len(free_gpus[node])
-                    node_with_min_moRE_gpUs = node
+                    node_with_min_more_GPUs = node
         if node_with_min_more_GPUs is not None:
             # only extracting the GPUs we need
             return (free_gpus[node_with_min_more_GPUs][:numGPUs_needed], True)

placement/placement.py

@@ -239,7 +239,7 @@ def _consolidated_placement(
                 # found a node with more GPUs then needed
                 if min_more_GPUs > len(free_gpus[node]):
                     min_more_GPUs = len(free_gpus[node])
-                    node_with_min_moRE_gpUs = node
+                    node_with_min_more_GPUs = node
         if node_with_min_more_GPUs is not None:
             # only extracting the GPUs we need
             return (free_gpus[node_with_min_more_GPUs][:numGPUs_needed], True)

schedulers/scheduler_policy.py

@@ -48,7 +48,7 @@ def schedule(
             gpu_df: Contains GPU dataframe.
 
         Returns:
-                "order_job" : Mandatory key, list of dicts of jobs in the
+                "order_job" : Mandatory key, list of dicts of jobs
                                  in the order they are supposed to run.
                 "run_all_jobs": Some scheduler will only output the jobs to
                                     run which will fit on the GPU or expecting