#!/usr/bin/env python # -*- coding: utf-8 -*- ''' 由于数据很大，测试使用部分数据！ 按照磁盘占用率从大到小装箱，即按照磁盘先用完为止进行分配实例到主机。 @Auther :liuyuqi.gov@msn.cn @Time :2018/7/7 0:43 @File :sort_by_disk.py ''' import matplotlib matplotlib.use('Agg') import pandas as pd import time import libs.save_result df1 = pd.read_csv("../data/scheduling_preliminary_app_resources_20180606.csv", encoding="utf-8") df3 = pd.read_csv("../data/test-instance.csv") # print(df3["cpu"].value_counts()) # print(df3.head()) df3["cpu"] = df3["cpu"].astype("float") df3["disk"] = df3["disk"].astype("float") df3["mem"] = df3["mem"].astype("float") df3["M"] = df3["M"].astype("float") df3["P"] = df3["P"].astype("float") df3["PM"] = df3["PM"].astype("float") df3["isdeploy"] = False # machine # 其实就两类，所以就不需要导入数据了。 # 限制表 df4 = pd.read_csv("../data/scheduling_preliminary_app_interference_20180606.csv", header=None, names=list(["appid1", "appid2", "max_interference"]), encoding="utf-8") result = pd.DataFrame(columns=list(["instanceid", "machineid"]), data=list()) tem_pre_disk = tem_pre_mem = tem_pre_cpu = tem_pre_P = tem_pre_M = tem_pre_PM = 0 tem_disk = tem_mem = tem_cpu = tem_P = tem_M = tem_PM = 0 tmp_stand_cpu1 = 32 tmp_stand_mem1 = 64 tmp_stand_disk1 = 600 tmp_stand_cpu2 = 92 tmp_stand_mem2 = 288 tmp_stand_disk2 = 600 tmp_stand_P = 7 tmp_stand_M1 = 3 tmp_stand_M2 = 7 tmp_stand_PM1 = 7 tmp_stand_PM2 = 9 machine_count = 0 # 3000小机器，3000大机器。所以在小机器用完换大机器 j = 1 # j表示主机序号，从1-3000，3001到6000 is_deploy = False # 主机j是否部署了instance deploy_list = list() # 主机j部署的instanceid实例 # 各app之间的限制 def restrictApps(instance, deploy_list): len_list = len(deploy_list) if len_list == 0: return True else: ct = pd.Series(deploy_list).value_counts() for k, v in ct.items(): tmp = df4.loc[(df4["appid1"] == k) & (df4["appid2"] == instance)] row, col = tmp.shape if row > 0: if ct[instance] + 1 > tmp["max_interference"]: return False else: # 在限制表中找不到限制条件 return True # 执行部署方案 def deploy(): global j, is_deploy, tem_mem, tem_cpu, tem_disk, tem_P, tem_M, tem_PM, tem_pre_disk, tem_pre_mem, \ tem_pre_cpu, tem_pre_P, tem_pre_M, tem_pre_PM, result, df3, deploy_list print("------------开始部署啦--------------") start = time.time() row, column = df3.shape while row > 0: deployInstance() # 整个instace都遍历了，第j主机无法再放入一个，所以添加j+1主机 df3 = df3[df3["isdeploy"] == False] row, column = df3.shape df3 = df3.reset_index(drop=True) j = j + 1 # j++之后表示新建主机，所以新主机没有部署任何实例，为false，然后初始化所有其他参数 is_deploy = False tem_pre_disk = tem_pre_mem = tem_pre_cpu = tem_pre_P = tem_pre_M = tem_pre_PM = 0 tem_disk = tem_mem = tem_cpu = tem_P = tem_M = tem_PM = 0 deploy_list = list() # 部署完事 print("------------部署完啦--------------") end = time.time() print("总共耗时：", end - start, "秒") print("总共需要主机数：", j) print("部署方案前几条示意：", result.head()) libs.save_result.save_result(result) def deployInstance(): ''' 根据限制部署实例到主机上 :param row: 根据剩余的instance数量循环 :param j: 第j台主机 :return: 暂未定返回值，None ''' global is_deploy, tem_mem, tem_cpu, tem_disk, tem_P, tem_M, tem_PM, tem_pre_disk, tem_pre_mem, tem_pre_cpu, tem_pre_P, tem_pre_M, tem_pre_PM, result, j, df3, deploy_list for row in df3.itertuples(): i = row.Index tem_pre_cpu = tem_cpu + row.cpu tem_pre_mem = tem_mem + row.mem tem_pre_disk = tem_disk + row.disk # 当前磁盘消耗 tem_pre_P = tem_P + row.P tem_pre_M = tem_M + row.M tem_pre_PM = tem_PM + row.PM # if 满足限制表条件，则把当前实例部署到这台主机上。 if j < 3000: # 使用小主机 if is_deploy == True: if tem_pre_disk < tmp_stand_disk1: # 磁盘够 if restrictApps(instance=row.instanceid, deploy_list=deploy_list): if tem_pre_mem < tmp_stand_mem1: # 内存够 if tem_pre_cpu < tmp_stand_cpu1: # CPU够 if tem_pre_M < tmp_stand_M1: if tem_pre_P < tmp_stand_P: if tem_pre_PM < tmp_stand_PM1: # 条件都满足，则把instance放入主机，同时df3表中去掉这个部署好的一行 result = result.append(pd.DataFrame( [{"instanceid": row.instanceid, "machineid": "machine_" + str(j)}])) tem_disk = tem_disk + row.disk tem_mem = tem_mem + row.mem tem_cpu = tem_cpu + row.cpu tem_P = tem_P + row.P tem_M = tem_M + row.M tem_PM = tem_PM + row.PM df3.loc[i, "isdeploy"] = True deploy_list.append(row.instanceid) else: # 主机j没有部署实例，则先部署一个 result = result.append( pd.DataFrame([{"instanceid": row.instanceid, "machineid": "machine_" + str(j)}])) tem_disk = tem_disk + row.disk tem_mem = tem_mem + row.mem tem_cpu = tem_cpu + row.cpu tem_P = tem_P + row.P tem_M = tem_M + row.M tem_PM = tem_PM + row.PM df3.loc[i, "isdeploy"] = True deploy_list.append(row.instanceid) # df3["isdeploy"][i] = True is_deploy = True else: # 使用大主机 if is_deploy == True: if tem_pre_disk < tmp_stand_disk2: # 磁盘够 if restrictApps(instance=row.instanceid, deploy_list=deploy_list): if tem_pre_mem < tmp_stand_mem2: # 内存够 if tem_pre_cpu < tmp_stand_cpu2: # CPU够 if tem_pre_M < tmp_stand_M2: if tem_pre_P < tmp_stand_P: if tem_pre_PM < tmp_stand_PM2: # 条件都满足，则把instance放入主机 result = result.append(pd.DataFrame( [{"instanceid": row.instanceid, "machineid": "machine_" + str(j)}])) tem_disk = tem_disk + row.disk tem_mem = tem_mem + row.mem tem_cpu = tem_cpu + row.cpu tem_P = tem_P + row.P tem_M = tem_M + row.M tem_PM = tem_PM + row.PM df3.loc[i, "isdeploy"] = True deploy_list.append(row.instanceid) else: # 主机j没有部署实例，则先部署一个 result = result.append( pd.DataFrame([{"instanceid": row.instanceid, "machineid": "machine_" + str(j)}])) tem_disk = tem_disk + row.disk tem_mem = tem_mem + row.mem tem_cpu = tem_cpu + row.cpu tem_P = tem_P + row.P tem_M = tem_M + row.M tem_PM = tem_PM + row.PM df3.loc[i, "isdeploy"] = True deploy_list.append(row.instanceid) is_deploy = True deploy()