流式数据处理

1、 直接登陆服务器：ssh 2014210***@thumedia.org -p 6349

创建streaming.py:   touch streaming.py,并且如下编辑：

<span style="font-size:14px;">#! /usr/bin/python
import logging
import math
import time
pg2count={}
t=1
while 1:
    fp=open('/tmp/hw3.log','r')
    for line in fp:
        line = line.strip()
        times, page, count = line.split()[0],line.split()[1],line.split()[2]
        if count.isdigit() & page.startswith('Page-'):
            try:
                pg2count[page] = [pg2count[page][0] + int(count),t]             
            except:
                pg2count[page] = [int(count),t]
    fp.close()
    a=sorted(pg2count.items(), key=lambda page:page[1][0], reverse = True)
    print '%s%s%s' % ('the page rank at current time ',times,' is:')
    for i in range(0,10):
        print '%s\t%d' % (a[i][0],a[i][1][0])
    logger = logging.getLogger()
    #set loghandler 
    file = logging.FileHandler("output.log")
    logger.addHandler(file)
    #set formater   
    formatter = logging.Formatter("%(asctime)s %(levelname)s %(message)s")
    file.setFormatter(formatter)
    #set log level 
    logger.setLevel(logging.NOTSET)
    logger.info('%s%s%s' % ('the page rank at current time ',times,' is:'))
    for i in range(0,10):
        logger.info('%s\t%d' % (a[i][0],a[i][1][0]))
        time.sleep(60)</span>

2、 写好代码之后测试运行：python streaming.py输出如下：

nohup: ignoring input and appending output to `nohup.out'，则表示后台运行成功，输出显示会保存到nohup.out中，

也可以查看output.log文件里的输出：

最后我们让它在后台一直执行：nohup python streaming.py &输出：

[1] 8994

2014210***@cluster-3-1:~$ nohup: ignoring input and appending output to `nohup.out'

一天之后，我们再次查看结果：

可以看到，累计的结果已经和第一次不太一样

3、 杀掉进程：ps -ef|grep 1020得到如下输出：

2014210***@cluster-3-1:~$ ps -ef|grep 1020

1020      7512  7471  0 Jan10 ?        00:00:00 sshd: 2014210***@pts/30

1020      7513  7512  0 Jan10 pts/30   00:00:00 -bash

1020      7574  7508  0 20:55 ?        00:00:00 sshd: 2014210***@pts/52

1020      7575  7574  0 20:55 pts/52   00:00:00 -bash

1020      8282  7575  0 21:04 pts/52   00:00:00 ps -ef

1020      8283  7575  0 21:04 pts/52   00:00:00 grep --color=auto 1020

1020      8994     1  0 13:20 ?        00:01:46 python streaming.py

1020     12260 12232  0 Jan10 ?        00:00:00 sshd: 2014210***@pts/35

1020     12261 12260  0 Jan10 pts/35   00:00:01 –bash

输入kill 8994：

2014210***@cluster-3-1:~$ kill 8994

2014210***@cluster-3-1:~$ ps -ef|grep 1020

1020      7512  7471  0 Jan10 ?        00:00:00 sshd: 2014210***@pts/30

1020      7513  7512  0 Jan10 pts/30   00:00:00 -bash

1020      7574  7508  0 20:55 ?        00:00:00 sshd: 2014210***@pts/52

1020      7575  7574  0 20:55 pts/52   00:00:00 -bash

1020      8335  7575  0 21:05 pts/52   00:00:00 ps -ef

1020      8336  7575  0 21:05 pts/52   00:00:00 grep --color=auto 1020

1020     12260 12232  0 Jan10 ?        00:00:00 sshd: 2014210***@pts/35

1020     12261 12260  0 Jan10 pts/35   00:00:01 –bash

至此，streaming.py运行结束。

 

Question

How can your design scale when the streaming is large and the calculation is complicated?

答：首先确定每个程序周期需要的时间，然后确定这段时间内的流数据能够保存在一块足够大的缓存区域，等到下个程序周期处理前一个缓存的流数据即可。