analysis-tool.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 166,
   "id": "b64351e2-2a42-451b-9901-fa88a694bd23",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 주요 라이브러리 \n",
    "import copy\n",
    "import re\n",
    "import json\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import glob\n",
    "import os\n",
    "import shutil\n",
    "from pathlib import Path\n",
    "throughput_interval = 0.1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "id": "043d5221-3fc1-4b9b-82d9-8df90ef128f6",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 커널 로그를 기반으로 패킷 송수신 내용 추출\n",
    "def read(file_name, throughput_analysis=True):\n",
    "    # 파일 읽기\n",
    "    with open(file_name) as json_file:\n",
    "        message = json_file.read()\n",
    "        \n",
    "    # 정규식 추출\n",
    "    # 예시: tcp_header_size,-544,507,48\n",
    "    # 데이터 이름, Flow id, 시간, 값\n",
    "    p = re.compile('''([a-z\\_0-9]+),([\\-0-9]+),([0-9]+),([0-9]+)''')\n",
    "    result = p.finditer(message)\n",
    "    \n",
    "    temp = []\n",
    "    for r in result:\n",
    "        key = str(r.group(1))\n",
    "        flow_id = int(r.group(2))\n",
    "        flow_time = int(r.group(3))\n",
    "        value = int(r.group(4))\n",
    "        temp.append([flow_time, flow_id, key, value])\n",
    "\n",
    "    # [time, flow_id, key, value] 형태의 테이블 생성\n",
    "    # 그러나 MPTCP가 아닌 다른 플로우 데이터도 함께 존재함\n",
    "    table = pd.DataFrame(temp, columns=['time', 'flow_id', 'key', 'value'])\n",
    "    \n",
    "    # 가장 많은 데이터를 보낸 3개의 플로우를 제외한 나머지 제거\n",
    "    table_not_included_meta_flow = table[table['flow_id'] != 0]\n",
    "    remove_id = table_not_included_meta_flow['flow_id'].value_counts().keys()[2:]\n",
    "    table = table.loc[table['flow_id'].isin(remove_id) == False]\n",
    "    \n",
    "    # 첫 번째 subflow 찾기 (=가장 처음으로 로그가 발생한 플로우)\n",
    "    temp = table[table['flow_id'] != 0] # MPTCP 레벨이 아닌 것 중\n",
    "    subflow_id = temp.sort_values([\"time\"], ascending=[True])\n",
    "    first_flow = subflow_id.iloc[0]['flow_id']\n",
    "    \n",
    "    # 각 flow별로 네임태그 지정\n",
    "    table.loc[table['flow_id'] == 0, 'flow_id'] = 'mptcp'\n",
    "    table.loc[table['flow_id'] == first_flow, 'flow_id'] = 'sub1'\n",
    "    table.loc[(table['flow_id'] != 'mptcp') & (table['flow_id'] != 'sub1') , 'flow_id'] = 'sub2'\n",
    "    \n",
    "    # 시간 값 정규화 (커널 시간 -> FLow 시작 시간)\n",
    "    min_time = table.loc[:,\"time\"].min() # 기준 값 설정\n",
    "    \n",
    "    # 250hz의 세분성\n",
    "    table.loc[:,\"time\"] = (table.loc[:,\"time\"] - min_time) / 250\n",
    "    \n",
    "    # total_bytes for mptcp level\n",
    "    flow = table.loc[(table['key'] == 'mptcp_send') & (table['value'] < 10000)].copy()\n",
    "    flow['flow_id'] = 'mptcp'\n",
    "    flow['key'] = 'send'\n",
    "    \n",
    "    flow2 = table.loc[(table['key'] == 'mptcp_send')].copy()\n",
    "    flow2['value'] = flow2['value'].apply(lambda x: x - 10000 if x >= 10000 else x)\n",
    "    flow2['flow_id'] = 'mptcp'\n",
    "    flow2['key'] = 'send_included_redundant'\n",
    "    table = table.loc[table['key'] != 'mptcp_send']\n",
    "    \n",
    "    table = pd.concat([table, flow, flow2])\n",
    "    table.name = file_name\n",
    "    \n",
    "    return statistics_table(table)\n",
    "    # Throughput 및 총 전송량 추가\n",
    "    total_throughput = []\n",
    "    \n",
    "    # range 함수는 int만 지원\n",
    "    bins = [i / 1000 for i in range(0, 60000, int(throughput_interval * 1000))]\n",
    "\n",
    "    option = []\n",
    "    for subflow in ['sub1', 'sub2']:\n",
    "        flow = select(table, subflow, \"send\")\n",
    "        transmission = flow.groupby(pd.cut(flow['time'], bins=bins)).sum()\n",
    "        y = [i / 1024 / 1024 for i in list(transmission['value'])]\n",
    "\n",
    "        flow = select(table, subflow, \"retransmission\")\n",
    "        retransmission = flow.groupby(pd.cut(flow['time'], bins=bins)).sum()\n",
    "\n",
    "        sent = 0\n",
    "        for i in range(0, len(list(transmission['value']))):\n",
    "            sent += transmission['value'].iloc[i]\n",
    "            option.append([bins[i], subflow, 'throughput', y[i] / throughput_interval])\n",
    "            option.append([bins[i], subflow, 'retransmission', retransmission['value'].iloc[i]])\n",
    "            option.append([bins[i], subflow, 'total_bytes', sent])\n",
    "    \n",
    "    # total_bytes for mptcp level\n",
    "    flow = table.loc[(table['key'] == 'mptcp_send') & (table['value'] >= 10000)].copy()\n",
    "    flow['value'] -= 10000\n",
    "    transmission_redundant = flow.groupby(pd.cut(flow['time'], bins=bins)).sum()\n",
    "    \n",
    "    flow = table.loc[(table['key'] == 'mptcp_send') & (table['value'] < 10000)]\n",
    "    transmission_real = flow.groupby(pd.cut(flow['time'], bins=bins)).sum()\n",
    "    \n",
    "    sent = 0\n",
    "    sent_included_redundant = 0\n",
    "    for i in range(0, len(list(transmission['value']))):\n",
    "        sent += transmission_real['value'].iloc[i]\n",
    "        sent_included_redundant += transmission_real['value'].iloc[i]\n",
    "        sent_included_redundant += transmission_redundant['value'].iloc[i]\n",
    "        option.append([bins[i], 'mptcp', 'total_bytes', sent])\n",
    "        option.append([bins[i], 'mptcp', 'total_bytes_included_redundant', sent_included_redundant])\n",
    "        \n",
    "    temp = pd.DataFrame(option, columns=['time', 'flow_id', 'key', 'value'])\n",
    "    table = pd.concat([table, temp])\n",
    "    table = table.loc[table['key'] != 'mptcp_send']\n",
    "    table.name = file_name\n",
    "    return statistics_table(table)\n",
    "\n",
    "# 정해진 간격으로 통계를 추출\n",
    "def statistics_table(table):\n",
    "    bins = [i / 1000 for i in range(0, 60000, int(throughput_interval * 1000))]\n",
    "    result = {}\n",
    "    for subflow in table[\"flow_id\"].unique():\n",
    "        result[subflow] = {}\n",
    "        table2 = table.loc[table[\"flow_id\"] == subflow]\n",
    "        for key in table2[\"key\"].unique():\n",
    "            table3 = table2.loc[table2['key'] == key, ['time', 'value']]\n",
    "            group = table3.groupby(pd.cut(table3['time'], bins=bins))['value']\n",
    "            new_table = pd.DataFrame()\n",
    "            new_table['time'] = bins[0:-1]\n",
    "            new_table['count'] = list(group.count())\n",
    "            new_table['min'] = list(group.min())\n",
    "            new_table['mean'] = list(group.mean())\n",
    "            new_table['max'] = list(group.max())\n",
    "            new_table['sum'] = list(group.sum())\n",
    "            new_table['cumsum'] = new_table['sum'].cumsum()\n",
    "            new_table['cumcount'] = new_table['count'].cumsum()\n",
    "            result[subflow][key] = new_table.dropna()\n",
    "    return result\n",
    "# 특정 subflow의 데이터를 선택하는 함수\n",
    "def select(table, subflow, key):\n",
    "    flow = table.loc[table[\"flow_id\"] == subflow]\n",
    "    flow = flow.loc[flow['key'] == key]\n",
    "    return flow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "id": "505b58f1-eff0-41cf-9c16-9e4bb7aee9a5",
   "metadata": {},
   "outputs": [],
   "source": [
    "red = read(\"logs\\\\그림 5\\\\redundant0.016 5MBps.txt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "id": "98719ddb-764b-46a5-bb7a-1c8a2dba380c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def view(table, flow_id, key, method, time_s = 0, time_e = 30, color=0, label=None, linestyle='-', marker=None, linewidth=1, linetype='a', ylabel=None, func=lambda x: x):\n",
    "    if (label == None): label = flow_id\n",
    "    flow = table[flow_id][key]\n",
    "    flow = flow.loc[(flow['time'] > time_s) & (flow['time'] < time_e)]\n",
    "    x = flow['time']\n",
    "    y = flow[method].apply(func)\n",
    "\n",
    "    nx = []\n",
    "    ny = []\n",
    "    for i in range(0,len(x)):\n",
    "        nx.append(x.iloc[i])\n",
    "        ny.append(y.iloc[i])\n",
    "        \n",
    "    colors = [['cornflowerblue', 'blue'],\n",
    "             ['sandybrown', 'coral'],\n",
    "             ['orchid', 'magenta'],\n",
    "             ['limegreen', 'forestgreen']]\n",
    "    if linetype == 'scatter':\n",
    "        plt.scatter(nx,ny, s=0.05, zorder=10, color=colors[color][1])\n",
    "    else:\n",
    "        plt.plot(nx, ny, linestyle, linewidth=linewidth,zorder=0, color=colors[color][0], label=label, marker=marker, markersize=2)\n",
    "    plt.xlabel(\"Flow time (sec)\")\n",
    "    if (ylabel == None):\n",
    "        plt.ylabel(\"\")\n",
    "    else:\n",
    "        plt.ylabel(ylabel)\n",
    "    plt.xlim(time_s, time_e)\n",
    "    #plt.ylim(0,500)\n",
    "    plt.title(flow_id + \" \" + key)\n",
    "# view(red, \"mptcp\", \"send_included_redundant\", \"sum\", func=lambda x: x / 1024 / 1024 / throughput_interval)\n",
    "# view(red, \"mptcp\", \"send\", \"sum\", func=lambda x: x / 1024 / 1024 / throughput_interval)\n",
    "view(red, \"mptcp\", \"rtt\", \"min\")\n",
    "view(red, \"mptcp\", \"rtt\", \"mean\")\n",
    "view(red, \"mptcp\", \"rtt\", \"max\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 175,
   "id": "6420f2de-401e-4d08-a449-f548d6487d78",
   "metadata": {},
   "outputs": [],
   "source": [
    "logs = {}\n",
    "for filename in Path('logs').rglob('*.txt'):\n",
    "    log_name = os.path.splitext(os.path.basename(filename))[0]\n",
    "    logs[str(filename)] = read(filename)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 181,
   "id": "afe27958-8bd3-41cc-b022-9094bd0f2764",
   "metadata": {},
   "outputs": [],
   "source": [
    "def save(table, path):\n",
    "    filename = path\n",
    "    base_path = os.path.splitext(filename)[0]\n",
    "    if os.path.isdir(base_path):\n",
    "        shutil.rmtree(base_path)\n",
    "        \n",
    "    os.mkdir(base_path)\n",
    "    for subflow in table:\n",
    "        for key in table[subflow]:\n",
    "            table[subflow][key].to_csv(base_path + \"/\" + subflow + \"_\" + key + \".csv\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "id": "411f9810-41ff-4630-b533-978d8bc6c2be",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "logs\\그림 3\\default.txt\n",
      "logs\\그림 4\\redundant.txt\n",
      "logs\\그림 5\\redundant0.008 5MBps.txt\n",
      "logs\\그림 5\\redundant0.016 5MBps.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\1MB 1초\\RTT강제변경_blest.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\1MB 1초\\RTT강제변경_default.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\1MB 1초\\RTT강제변경_only_fast.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\1MB 1초\\RTT강제변경_redundant.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\대역폭 제한 없음\\RTT강제변경_blest.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\대역폭 제한 없음\\RTT강제변경_default.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\대역폭 제한 없음\\RTT강제변경_only_fast.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\대역폭 제한 없음\\RTT강제변경_redundant.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\클라우드\\RTT강제변경_blest.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\클라우드\\RTT강제변경_default.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\클라우드\\RTT강제변경_only_fast.txt\n",
      "logs\\전체 실험\\1. RTT 강제 변경\\클라우드\\RTT강제변경_redundant.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\1MB 1초\\트래픽_blest.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\1MB 1초\\트래픽_default.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\1MB 1초\\트래픽_only_fast.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\1MB 1초\\트래픽_redundant.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\대역폭 제한 없음\\트래픽_blest.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\대역폭 제한 없음\\트래픽_default.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\대역폭 제한 없음\\트래픽_only_fast.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\대역폭 제한 없음\\트래픽_redundant.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\클라우드\\트래픽_blest.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\클라우드\\트래픽_default.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\클라우드\\트래픽_only_fast.txt\n",
      "logs\\전체 실험\\2. 트래픽이 발생함\\클라우드\\트래픽_redundant.txt\n",
      "logs\\전체 실험\\3. mmwave\\1MB 1초\\mmwave_blest.txt\n",
      "logs\\전체 실험\\3. mmwave\\1MB 1초\\mmwave_default.txt\n",
      "logs\\전체 실험\\3. mmwave\\1MB 1초\\mmwave_only_fast.txt\n",
      "logs\\전체 실험\\3. mmwave\\1MB 1초\\mmwave_redundant.txt\n",
      "logs\\전체 실험\\3. mmwave\\대역폭 제한 없음\\mmwave_blest.txt\n",
      "logs\\전체 실험\\3. mmwave\\대역폭 제한 없음\\mmwave_default.txt\n",
      "logs\\전체 실험\\3. mmwave\\대역폭 제한 없음\\mmwave_only_fast.txt\n",
      "logs\\전체 실험\\3. mmwave\\대역폭 제한 없음\\mmwave_redundant.txt\n",
      "logs\\전체 실험\\3. mmwave\\클라우드\\mmwave_blest.txt\n",
      "logs\\전체 실험\\3. mmwave\\클라우드\\mmwave_default.txt\n",
      "logs\\전체 실험\\3. mmwave\\클라우드\\mmwave_only_fast.txt\n",
      "logs\\전체 실험\\3. mmwave\\클라우드\\mmwave_redundant.txt\n"
     ]
    }
   ],
   "source": [
    "for i in logs:\n",
    "    print(i)\n",
    "    save(logs[i], i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2463cb24-137d-4883-9c01-1ec22ef0cab5",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}