add expected results and plotting

Author: qzweng

experiments/README.md

@@ -3,7 +3,7 @@
 ### Generate Run Scripts
 
 ```bash
-cd kubernetes-scheduler-simulator/experiments
+# pwd: kubernetes-scheduler-simulator/experiments
 python run_scripts/generate_run_scripts.py > run_scripts/run_scripts_0511.sh
 ```
 
@@ -33,15 +33,21 @@ cat experiments/run_scripts/run_scripts_0511.sh | while read i; do printf "%q\n"
 ```bash
 # pwd: kubernetes-scheduler-simulator/experiments
 ln -s 2023_0511 data # softlink it to data
+# pwd: kubernetes-scheduler-simulator/experiments/analysis
 cd analysis
-bash merge_bash.sh
+bash bash_merge.sh
+# The output was generated under "analysis_results"
+# The results of our large-scale experiments are cached in "expected_results" for comparison
 ```
 
 ### Plot
 
 ```bash
+# pwd: kubernetes-scheduler-simulator/experiments/analysis
+cd ..
 # pwd: kubernetes-scheduler-simulator/experiments
-ln -s analysis/analysis_results/* plot/ # softlink all csv under analysis_results/ to plot/
+cp analysis/analysis_results/* plot/ # copy all csv under analysis_results/ to plot/ for analysis
+# cp analysis/expected_results/* plot/ # if skipping the experiments and directly reuse our expected results for plotting
 cd plot
 python plot_paib_alloc.py              # Fig. 9(a)
 python plot_paib_frag_amount.py        # Fig. 7(a)

experiments/analysis/merge_bash.sh renamed to experiments/analysis/bash_merge.sh

@@ -1,9 +1,5 @@
 #!/bin/bash
-DATE="2023_0511"
-TRACE="openb_pod_list_default"
-ln -s "../${DATE}/${TRACE}" ../data/
-
 python3 merge_alloc_discrete_0920.py &
 python3 merge_frag_discrete_0928.py &
 python3 merge_frag_ratio_discrete_0928.py &
-python3 merge_fail_pods_1010.py &
+python3 merge_fail_pods_1010.py &

experiments/analysis/expected_results/analysis_allo_discrete.csv

@@ -1,11 +1,15 @@
+import os
 import pandas as pd
 from pathlib import Path
-from utils import move_tag_to_new_column
+
+RESULTDIR="analysis_results"
+DATADIR="data"
 RESULTDIR="analysis_results"
 DATADIR="data"
-root = Path(__file__).parents[1] # 0524
+filepath = os.path.abspath(__file__)
+root = Path(filepath).parents[1] # 0524
 data = root / DATADIR
-analysis = Path(__file__).parent # 0830
+analysis = Path(filepath).parent # 0830
 resultDir = analysis / RESULTDIR
 # .
 # ├── README.md
@@ -23,6 +27,47 @@
 # │   │   │   │   ├── [d] 42
 # ......
 
+TAG_SNAKE_LIST = ['init_schedule', 'post_eviction', 'post_deschedule', 'schedule_inflation', 'deschedule_inflation']
+def move_tag_to_new_column(df, tag_list=TAG_SNAKE_LIST):
+    meta_col = []
+    data_col = []
+    for col in df.columns:
+        is_data_col = False
+        for tag in tag_list:
+            if col.endswith("_" + tag):
+                data_col.append(col)
+                is_data_col = True
+                break
+        if is_data_col == False:
+            meta_col.append(col)
+    
+    out_row_list = []
+    for _, row in df.iterrows():
+        orig_dict = dict(row)
+        meta_dict = {}
+        for col in meta_col:
+        # for col in NONTAG_COL:
+            if col in orig_dict:
+                meta_dict[col] = orig_dict[col]
+        # print("meta_dict:", meta_dict)
+
+        for tag in tag_list:
+            data_dict = {}
+            data_dict.update(meta_dict)
+            data_dict['tag'] = tag
+            found = 0
+            for col in data_col:
+                if col.endswith("_" + tag):
+                    key = col[:-(len(tag)+1)]
+                    # print(tag, '+', key,'=',col)
+                    data_dict[key] = orig_dict.get(col)
+                    found = 1
+            if found == 1:
+                # print("data_dict:", data_dict)
+                data_row = pd.DataFrame().from_dict(data_dict, orient='index').T
+                out_row_list.append(data_row)
+    return pd.concat(out_row_list)
+
 def exit_and_save_to_csv(dflist):
     dfo = pd.concat(dflist)
     resultDir.mkdir(exist_ok=True)

experiments/analysis/expected_results/analysis_fail.csv

@@ -1,14 +1,15 @@
+import os
 import pandas as pd
 from pathlib import Path
-from utils import get_total_num_gpu
 
 DATADIR="data"
 IN_FILE="analysis_fail.out"
 RESULTDIR="analysis_results"
 OUT_CSV="analysis_fail.csv"
-root = Path(__file__).parents[1] # 0524
+filepath = os.path.abspath(__file__)
+root = Path(filepath).parents[1] # 0524
 data = root / DATADIR
-analysis = Path(__file__).parent # 0830
+analysis = Path(filepath).parent # 0830
 resultDir = analysis / RESULTDIR
 # .
 # ├── README.md
@@ -25,6 +26,8 @@
 # │   │   │   ├── [d] 0.6
 # │   │   │   │   ├── [d] 42
 # ......
+def get_total_num_gpu(workload):
+    return 6212
 
 def parse_fail_out_file_to_df(ifile):
     with open(ifile, 'r') as f:

experiments/analysis/expected_results/analysis_frag_discrete.csv

@@ -1,12 +1,14 @@
+import os
 import pandas as pd
 from pathlib import Path
-from utils import get_total_num_gpu
 
 RESULTDIR="analysis_results"
 DATADIR="data"
-root = Path(__file__).parents[1] # 0524
+
+filepath = os.path.abspath(__file__)
+root = Path(filepath).parents[1] # 0524
 data = root / DATADIR
-analysis = Path(__file__).parent # 0830
+analysis = Path(filepath).parent # 0830
 resultDir = analysis / RESULTDIR
 # .
 # ├── README.md
@@ -23,6 +25,8 @@
 # │   │   │   ├── [d] 0.6
 # │   │   │   │   ├── [d] 42
 # ......
+def get_total_num_gpu(workload):
+    return 6212
 
 def exit_and_save_to_csv(dflist):
     dfo = pd.concat(dflist)

experiments/analysis/expected_results/analysis_frag_ratio_discrete.csv

@@ -1,12 +1,13 @@
+import os
 import pandas as pd
 from pathlib import Path
-from utils import get_total_num_gpu
 
 RESULTDIR="analysis_results"
 DATADIR="data"
-root = Path(__file__).parents[1] # 0524
+filepath = os.path.abspath(__file__)
+root = Path(filepath).parents[1] # 0524
 data = root / DATADIR
-analysis = Path(__file__).parent # 0830
+analysis = Path(filepath).parent # 0830
 resultDir = analysis / RESULTDIR
 # .
 # ├── README.md
@@ -24,6 +25,9 @@
 # │   │   │   │   ├── [d] 42
 # ......
 
+def get_total_num_gpu(workload):
+    return 6212
+
 def exit_and_save_to_csv(dflist):
     dfo = pd.concat(dflist)
     resultDir.mkdir(exist_ok=True)