File size: 10,646 Bytes
fb0bedf |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 |
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Process Utilization Files and Generate a Pickle Files for Plotting"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Note that running following scripts is time consuming and should be done only once. The pickle files are provided in the `util_pkl` directory."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import pytz\n",
"import pickle\n",
"import numpy as np\n",
"import pandas as pd\n",
"\n",
"from glob import glob\n",
"from pathlib import Path\n",
"from utils import cluster_metric_header, dcgm_metric_header\n",
"\n",
"\"\"\" cluster_metric_header\n",
"0: \"CPU\",\n",
"1: \"MEMORY\",\n",
"2: \"IB_SEND\",\n",
"3: \"RECEIVE\",\n",
"\"\"\"\n",
"\"\"\" dcgm_metric_header\n",
"0: XID_ERRORS\n",
"1: GPU_TEMP\n",
"2: MEMORY_TEMP\n",
"3: MEM_CLOCK\n",
"4: MEM_COPY_UTIL\n",
"5: FB_FREE\n",
"6: FB_USED\n",
"7: DRAM_ACTIVE\n",
"8: POWER_USAGE\n",
"9: GPU_UTIL\n",
"10: PIPE_TENSOR_ACTIVE\n",
"11: SM_ACTIVE\n",
"12: SM_OCCUPANCY\n",
"\"\"\"\n",
"\n",
"SAVEPKL = \"./util_pkl\"\n",
"\n",
"\n",
"def read_csv_with_concat(path=\"./csv\", file_name=None):\n",
" file = Path(path, f\"{file_name}.csv\")\n",
"\n",
" if file.exists():\n",
" # If original file exists, read it directly\n",
" df = pd.read_csv(file)\n",
" print(f\"Reading {file_name}\")\n",
" else:\n",
" # If original file does not exist, read all the split files\n",
" split_files = sorted(glob(f\"{path}/{file_name}-2023-*.csv\"))\n",
" print(f\"Reading splitted files: {split_files}\")\n",
" df = pd.concat([pd.read_csv(split_file) for split_file in split_files])\n",
" df.reset_index(drop=True, inplace=True)\n",
" return df\n",
"\n",
"\n",
"def read_concat_parse_save_cluster_metrics(path=\"./csv_cpu\", metrics=cluster_metric_header):\n",
" for metric in metrics:\n",
" data = read_csv_with_concat(path=path, file_name=metric)\n",
" data.drop_duplicates(subset=[\"Time\"], inplace=True)\n",
" data.sort_values(by=\"Time\", inplace=True)\n",
" data[\"Time\"] = pd.to_datetime(data[\"Time\"], unit=\"s\").dt.tz_localize(pytz.utc).dt.tz_convert(\"Asia/Shanghai\")\n",
" data.set_index(\"Time\", drop=True, inplace=True)\n",
" print(f\"Column Number: {len(list(data.columns))}, {len(set(list(data.columns)))}\")\n",
"\n",
" if \"NODE_MEMORY\" in metric:\n",
" # Around 2 hours has some bug (ip has additional '.1', like '10.140.0.131' -> '10.140.0.131.1')\n",
" data = data[(data.index < \"2023-07-19 11:35:00\") | (data.index > \"2023-07-19 14:01:00\")]\n",
"\n",
" if \"NODE_CPU\" in metric or \"NODE_MEMORY\" in metric:\n",
" data = data * 100 # CPU / Memory Utilization (%)\n",
"\n",
" if \"NODE_IB\" in metric:\n",
" data.rename(columns=lambda x: x.replace(\"-mlx5_0\", \"\"), inplace=True) # Simplified, since one IB NIC per server\n",
"\n",
" data.dropna(axis=1, how=\"all\", inplace=True)\n",
" data = data.round(3)\n",
" data.to_csv(f\"./{metric}.csv\")\n",
"\n",
"\n",
"def read_concat_parse_save_dcgm_metrics(path=\"./csv\", metrics=dcgm_metric_header):\n",
" for metric in metrics:\n",
" data = read_csv_with_concat(path=path, file_name=metric)\n",
" data.drop_duplicates(subset=[\"Time\"], inplace=True)\n",
" data.sort_values(by=\"Time\", inplace=True)\n",
" data[\"Time\"] = pd.to_datetime(data[\"Time\"], unit=\"s\").dt.tz_localize(pytz.utc).dt.tz_convert(\"Asia/Shanghai\")\n",
" data.set_index(\"Time\", drop=True, inplace=True)\n",
" print(f\"Column Number: {len(list(data.columns))}, {len(set(list(data.columns)))}\")\n",
"\n",
" # if \"XID\" in metric or \"TEMP\" in metric or \"CLOC\" in metric:\n",
" # data = data.astype(int, errors='ignore')\n",
"\n",
" if \"ACTIVE\" in metric or \"OCCUPANCY\" in metric:\n",
" data = data * 100 # CPU / Memory Utilization (%)\n",
" data = data.round(3)\n",
"\n",
" if \"POWER\" in metric:\n",
" data = data.round(1)\n",
"\n",
" data.dropna(axis=0, how=\"all\", inplace=True)\n",
" data.dropna(axis=1, how=\"all\", inplace=True)\n",
" data.to_csv(f\"./{metric}.csv\")\n",
"\n",
"\n",
"def calculate_sum_cdf_axis100(df, dot_num=1000):\n",
" \"\"\"\n",
" Calculate quantity percentile CDF, y-axis: 0-100%,\n",
" \"\"\"\n",
" print(\"Parsing\")\n",
" data = df.melt(id_vars=\"Time\", var_name=\"Server\")\n",
" data.dropna(subset=[\"value\"], inplace=True)\n",
"\n",
" y = np.linspace(0, 1, num=dot_num)\n",
" x = data[\"value\"].quantile(y).values\n",
" y = y * 100\n",
" return x, y\n",
"\n",
"\n",
"def calculate_num_cdf_axis100(df, dot_num=1000):\n",
" \"\"\"\n",
" Calculate quantity percentile CDF, y-axis: 0-100%,\n",
" \"\"\"\n",
" print(\"Parsing\")\n",
" data = df.melt(id_vars=\"Time\", var_name=\"Server\")\n",
" data.dropna(subset=[\"value\"], inplace=True)\n",
" # data.sort_values('value', ascending=True, inplace=True)\n",
" # data.reset_index(drop=True, inplace=True)\n",
"\n",
" y = np.linspace(0, 1, num=dot_num)\n",
" x = data[\"value\"].quantile(y).values\n",
" y = y * 100\n",
" return x, y"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Example 1: Prometheus Metics (e.g., CPU and Memory Utilization)\n",
"\n",
"You can change to any metric you want to plot by changing the `file_name` variable in the following script."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"data_cpu = read_csv_with_concat(path=\"./seren\", file_name=cluster_metric_header[0])\n",
"data_mem = read_csv_with_concat(path=\"./seren\", file_name=cluster_metric_header[1])\n",
"x1, y1 = calculate_num_cdf_axis100(data_cpu)\n",
"x2, y2 = calculate_num_cdf_axis100(data_mem)\n",
"print(\n",
" f'CPU Period: (Start) {data_cpu.at[0, \"Time\"].split(\":\")[0]}h (End) {data_cpu.at[len(data_cpu)-1, \"Time\"].split(\":\")[0]}h'\n",
")\n",
"print(\n",
" f'MEM Period: (Start) {data_mem.at[0, \"Time\"].split(\":\")[0]}h (End) {data_mem.at[len(data_mem)-1, \"Time\"].split(\":\")[0]}h'\n",
")\n",
"\n",
"with open(f\"{SAVEPKL}/util_cpu_mem_seren.pkl\", \"wb\") as file:\n",
" pickle.dump([x1, y1, x2, y2], file)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Example 2: NVIDIA DCGM Metics (e.g., GPU and GPU Memory Utilization)\n",
"\n",
"You can change to any metric you want to plot by changing the `file_name` variable in the following script."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"data_gpu_util = read_csv_with_concat(path=\"./seren\", file_name=dcgm_metric_header[9]) # \"DCGM_FI_DEV_GPU_UTIL\"\n",
"data_gpu_mem = read_csv_with_concat(path=\"./seren\", file_name=dcgm_metric_header[6]) # \"DCGM_FI_DEV_FB_USED\"\n",
"data_gpu_mem.iloc[:, 1:] = 100 * data_gpu_mem.iloc[:, 1:] / (80 * 1024)\n",
"x1, y1 = calculate_num_cdf_axis100(data_gpu_util)\n",
"x2, y2 = calculate_num_cdf_axis100(data_gpu_mem)\n",
"\n",
"with open(f\"{SAVEPKL}/util_gpu_util_mem_seren.pkl\", \"wb\") as file:\n",
" pickle.dump([x1, y1, x2, y2], file)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Processing IPMI Power Files"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"df_AB = pd.read_csv(\"./ipmi/GPU_AB_Power.csv\", parse_dates=[\"Time\"])\n",
"df_C = pd.read_csv(\"./ipmi/GPU_C_Power.csv\", parse_dates=[\"Time\"])\n",
"df_D = pd.read_csv(\"./ipmi/CPU_D_Power.csv\", parse_dates=[\"Time\"])\n",
"\n",
"df_A = df_AB.dropna()\n",
"df_B = df_AB[df_AB.isna().any(axis=1)] # Type B without MEM_Power record\n",
"\n",
"dfs = {\"GPU_A\": df_A, \"GPU_B\": df_B, \"GPU_C\": df_C, \"CPU_D\": df_D}\n",
"\n",
"# Extract sys_total_power\n",
"df_A_power = df_A[[\"Time\", \"Sys_Total_Power\"]]\n",
"df_B_power = df_B[[\"Time\", \"Sys_Total_Power\"]]\n",
"df_C_power = df_C[[\"Time\", \"Sys_Total_Power\"]]\n",
"df_gpu = pd.concat([df_A_power, df_B_power, df_C_power])\n",
"\n",
"x1, y1 = calculate_sum_cdf_axis100(df_gpu)\n",
"x2, y2 = calculate_sum_cdf_axis100(df_D[[\"Time\", \"Sys_Total_Power\"]])\n",
"\n",
"with open(f\"./server_power.pkl\", \"wb\") as file:\n",
" pickle.dump([x1, y1, x2, y2], file)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Processing Philly GPU Utilization Data"
]
},
{
"cell_type": "code",
"execution_count": 45,
"metadata": {},
"outputs": [],
"source": [
"philly = pd.read_csv(\n",
" \"./philly/philly_gpu_util.csv\", on_bad_lines=\"skip\", header=0\n",
") # Please refer to their official repo for the data\n",
"cols = list(philly.columns)\n",
"philly = philly.drop(columns=[cols[-1]])\n",
"philly.reset_index(inplace=True)\n",
"philly.columns = cols\n",
"philly.rename(columns={\"time\": \"Time\"}, inplace=True)\n",
"philly = philly.drop(columns=[cols[1]])\n",
"\n",
"x1, y1 = calculate_num_cdf_axis100(philly)\n",
"with open(f\"{SAVEPKL}/util_gpu_util_philly.pkl\", \"wb\") as file:\n",
" pickle.dump([x1, y1], file)"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "base",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.16"
},
"orig_nbformat": 4
},
"nbformat": 4,
"nbformat_minor": 2
}
|