首页 > 代码库 > 从GoogleClusterData统计每个用户的使用率、平均每次出价
从GoogleClusterData统计每个用户的使用率、平均每次出价
之前将google cluster data导入了Azure上的MySQL数据库,下一步就是对这些数据进行分析,
挖掘用户的使用规律了。
首先,为了加快执行速度,对user,time等加入索引。
然后就可以使用以下代码进行统计了。
import osimport MySQLdbimport timeimport threaddef use4ADay(day, users): conn=MySQLdb.connect(host="localhost",user="root",passwd="123456",db="googleclusterdata",charset="utf8") cursor = conn.cursor() msAday = 24*60*60*1000000 for user in users: user = user[0] print user use4ADay.user = user print ‘day %s‘ %day startTime = (day - 1) * msAday endTime = day * msAday dayCPUUse = 0 dayMEMUse = 0 dayDiskUse = 0 order = "select job_id from job_events where time >= %s and time < %s and user = ‘%s‘" %(startTime, endTime, user) print order cursor.execute(order) job_ids = cursor.fetchall() for job_id in job_ids: job_id = job_id[0] print ‘day %s‘ %day order = "select task_index, event_type, cpu_request, memory_request, disk_space_request, time from task_events where time >= %s and time < %s and job_id = %d order by task_index" %(startTime, endTime, job_id) print order cursor.execute(order) tasks = cursor.fetchall() print ‘tasks get‘ i = 0 while i < len(tasks) - 1: task = tasks[i] if task[1] == 1: task_index = task[0] nextEvent = tasks[i+1] if (nextEvent[1] == 4 or nextEvent[1] == 5) and nextEvent[0] == task_index: taskLife = (nextEvent[5] - tasks[i][5]) / (10.0**6) dayCPUUse += taskLife * task[2] dayMEMUse += taskLife * task[3] dayDiskUse += taskLife * task[4] #print ‘task: ‘, task_index, dayCPUUse, dayMEMUse, dayDiskUse i = i+1 #print ‘job: ‘, job_id, dayCPUUse, dayMEMUse, dayDiskUse fOut = open(‘C:\\userUsageEachDay\\day%d.txt‘ %day, ‘a‘) fOut.write(‘%s\t%f\t%f\t%f\n‘ %(user, dayCPUUse, dayMEMUse, dayDiskUse)) fOut.close() print ‘day %d finish‘ %day conn.close() conn=MySQLdb.connect(host="localhost",user="root",passwd="123456",db="googleclusterdata",charset="utf8")cursor = conn.cursor()#get all user_nameorder = "select distinct user from job_events"print ordercursor.execute(order)users = cursor.fetchall()conn.close()for day in range(1, 30): try: use4ADay(day, users) except: print ‘day‘, day, ‘failed!!‘ fOut = open(‘C:\\failed.txt‘, ‘a‘) fOut.write(‘%s\t%d\t\n‘ %(use4ADay.user, day)) fOut.close() #print ‘starting thread for day %d‘ %day #thread.start_new_thread(use4ADay, (day, users, ) )#use4ADay(2, users)
下一步,是统计每个用户整个月的消费频率,以及每次消费的平均消费量
fDay1 = open(‘C:\\Usage\\day1.txt‘)users = []for l in fDay1.readlines(): l = l.split(‘\t‘) user = l[0] users.append(user)fDay1.close()#fOut = open(‘C:\\UseTraceOfAllUsers.txt‘, ‘w‘)for user in users: useDays = 0 allPrice = 0 for day in range(1,30): f = open(‘C:\\Usage\\day%d.txt‘ %day) isFind = False for l in f.readlines(): if l.count(user) > 0: l = l.strip() l = l.split(‘\t‘) cpu = float(l[1]) mem = float(l[2]) disk = float(l[3]) money = 1.92*cpu + 15.6*mem + 1.2*disk assert(money>=0) isFind = True break if isFind and money != 0: useDays += 1 allPrice += money f.close() if useDays != 0: pass #fOut.write(‘%s\t%s\n‘ %(str(useDays/29.0), str(allPrice/useDays)))fOut.close()
最后就可以使用matlab进行画图啦。
x = load(‘C:\UseTraceOfAllUsers.txt‘)plot(x(:,1), x(:,2), ‘o‘);
结果如下:
对平均使用量取个对数的话
x = load(‘C:\UseTraceOfAllUsers.txt‘)plot(x(:,1), log(x(:,2)), ‘o‘);
从GoogleClusterData统计每个用户的使用率、平均每次出价
声明:以上内容来自用户投稿及互联网公开渠道收集整理发布,本网站不拥有所有权,未作人工编辑处理,也不承担相关法律责任,若内容有误或涉及侵权可进行投诉: 投诉/举报 工作人员会在5个工作日内联系你,一经查实,本站将立刻删除涉嫌侵权内容。