diff --git a/cernet/stat.py b/cernet/stat.py
index 66fcc41..6843ce9 100644
--- a/cernet/stat.py
+++ b/cernet/stat.py
@@ -9,7 +9,6 @@
 import sys
 from operator import add
 import re
-import csv
 import urllib
 import os
 
@@ -50,34 +49,13 @@ def cleanse(record, ipdict, parser, air):
 def stat(x, prefix, features, f):
     if len(features) == 1:
         fea = f[features[0]]
-        result = x.map(lambda x: (x.split('|')[fea], 1)).reduceByKey(add).collect()
+        result = x.map(lambda x: (x.split('|')[fea], 1)).reduceByKey(add)
     else:
-        result = stat1(x, features, f).collect()
-    filename = 'result/[%s]%s.csv'%(prefix, '_'.join(features))
-    output(result, filename, len(features))
-
-def output(result, filename, n):
-    if len(result) == 0:
+        result = stat1(x, features, f)
+    filename = 'result/[%s]%s'%(prefix, '_'.join(features))
+    if result.count() == 0:
         return
-    with open(filename, 'ab+') as final:
-        spamwriter = csv.writer(final, dialect='excel')
-        if n-1:
-            col= result[0][1]
-            for line in result[1:]:
-                if len(line[1]) > len(col):
-                    col = line[1]
-            spamwriter.writerow(['']+list(col))
-            for line in result:
-                row = [line[0]]
-                for item in col:
-                    try:
-                        row.append(line[1][item])
-                    except KeyError:
-                        row.append(0)
-                spamwriter.writerow(row)
-        else:
-            for line in result:
-                spamwriter.writerow(list(line))
+    result.repartition(1).saveAsTextFile(filename)
 
 def stat1(x, features, f):
     feas = []
@@ -93,7 +71,11 @@ def stat2(x, n):
             d2[item.split('|')[n]] = d2[item.split('|')[n]] + 1
         except KeyError:
             d2[item.split('|')[n]] = 1
-    return (d1, d2)
+    l2 = list(sorted(list(d2)))
+    tmp = ''
+    for item in l2:
+        tmp = tmp + '%s$%s|'%(item, d2[item])
+    return (d1, tmp)
 
 def main(sc, path):
     os.system('mkdir result')
@@ -177,11 +159,11 @@ def main(sc, path):
     for dim in f:
         stat(cernet2, 'total', [dim], f)
     stat(cernet2, 'total', ['month', 'hour'], f)
-    stat(cernet2, 'total', ['day', 'hour'], f)
-    stat(cernet_trip, 'trip',  ['keyword', 'month'], f)
-    stat(cernet_trip, 'trip',  ['keyword', 'school'], f)
-    stat(cernet_shop, 'shop', ['keyword', 'hour'], f)
-    stat(cernet_shop, 'shop',  ['keyword', 'month'], f)
+    stat(cernet2, 'total', ['hour', 'day'], f)
+    stat(cernet_trip, 'trip',  ['month', 'keyword'], f)
+    stat(cernet_shop, 'shop',  ['month', 'keyword'], f)
+    stat(cernet_shop, 'shop', ['hour', 'keyword'], f)
+    stat(cernet_shop, 'shop',  ['day', 'keyword'], f)
     stat(cernet_shop, 'shop', ['host', 'hour'], f)
     stat(cernet_shop, 'shop', ['host', 'month'], f)
     stat(cernet_shop, 'shop', ['host', 'day'], f)
@@ -195,5 +177,7 @@ def main(sc, path):
 if __name__ == "__main__":
     path = sys.argv[1]
     conf = SparkConf().setAppName(APP_NAME)
+    conf = conf.set("spark.akka.frameSize", "1000")
+    conf = conf.set("spark.shuffle.consolidateFiles", "True")
     sc = SparkContext(conf=conf)
     main(sc, path)
\ No newline at end of file