我建议这样设计：
mapper从文件中获取一行并输出ip作为密钥和一对网站，1作为值
合路器和减速器。获取ip作为密钥和一系列（website，count）对，按网站聚合它们（使用hashmap），并输出ip、website和count作为输出。
实现这一点需要您实现自定义可写来处理一对。
就我个人而言，我会用spark来做这件事，除非你太在意我的表现。有了pyspark，就这么简单了：

rdd = sc.textFile('/sparkdemo/log.txt')
counts = rdd.map(lambda line: line.split()).map(lambda line: ((line[0], line[1]), 1)).reduceByKey(lambda x, y: x+y)
result = counts.map(lambda ((ip, url), cnt): (ip, (url, cnt))).groupByKey().collect()
for x in result:
    print 'IP: %s' % x[0]
    for w in x[1]:
        print '    website: %s count: %d' % (w[0], w[1])

您的示例的输出是：

IP: 192.168.72.224
    website: www.facebook.com count: 2
    website: www.m4maths.com count: 2
    website: www.google.com count: 5
    website: www.gmail.com count: 4
    website: www.indiabix.com count: 8
    website: www.yahoo.com count: 3
IP: 192.168.72.177
    website: www.yahoo.com count: 14
    website: www.google.com count: 3
    website: www.facebook.com count: 3
    website: www.m4maths.com count: 3
    website: www.indiabix.com count: 1
IP: 192.168.198.92
    website: www.facebook.com count: 4
    website: www.m4maths.com count: 3
    website: www.yahoo.com count: 3
    website: www.askubuntu.com count: 2
    website: www.indiabix.com count: 1
    website: www.google.com count: 5
    website: www.gmail.com count: 1

我用java编写了相同的逻辑

public class UrlHitMapper extends Mapper<Object, Text, Text, Text>{
    public void map(Object key, Text value, Context contex) throws IOException, InterruptedException {
        System.out.println(value);
        StringTokenizer st=new StringTokenizer(value.toString());
        if(st.hasMoreTokens())
            contex.write(new Text(st.nextToken()), new Text(st.nextToken()));
    }
}
public class UrlHitReducer extends Reducer<Text, Text, Text, Text>{
    public void reduce(Text key, Iterable<Text> values, Context context)
            throws IOException, InterruptedException {
        HashMap<String, Integer> urlCount=new HashMap<>();
        String url=null;
        Iterator<Text> it=values.iterator();
        while (it.hasNext()) {
            url=it.next().toString();
            if(urlCount.get(url)==null)
                urlCount.put(url, 1);
            else
                urlCount.put(url,urlCount.get(url)+1);
        }
        for(Entry<String, Integer> k:urlCount.entrySet())
        context.write(key, new Text(k.getKey()+"    "+k.getValue()));
    }
}
public class UrlHitCount extends Configured implements Tool {
    public static void main(String[] args) throws Exception {
        ToolRunner.run(new Configuration(), new UrlHitCount(), args);
    }
    public int run(String[] arg0) throws Exception {
        Job job = Job.getInstance(getConf());
        job.setJobName("url-hit-count");
        job.setOutputKeyClass(Text.class);
        job.setOutputValueClass(Text.class);
        job.setMapperClass(UrlHitMapper.class);
        job.setReducerClass(UrlHitReducer.class);  
        job.setOutputFormatClass(TextOutputFormat.class);
        FileInputFormat.setInputPaths(job, new Path("input/urls"));
        FileOutputFormat.setOutputPath(job, new Path("url_otput"+System.currentTimeMillis()));
        job.setJarByClass(WordCount.class);
        job.submit();
        return 1;
    }
}