github.com/lnzx/goofys@v0.24.0/bench/bench_format.py (about)

     1  #!/usr/bin/python
     2  
     3  import numpy
     4  import sys
     5  
     6  def filter_outliers(numbers, mean, std):
     7      if len(numbers) == 1:
     8          return numbers
     9      return filter(lambda x: abs(x - mean) < 2 * std, numbers)
    10  
    11  op_str = {
    12      'create_files' : 'Create 100 files',
    13      'create_files_parallel' : 'Create 100 files (parallel)',
    14      'rm_files' : 'Unlink 100 files',
    15      'rm_files_parallel' : 'Unlink 100 files (parallel)',
    16      'ls_files' : 'ls with 1000 files',
    17      'find_files' : "`find' with 1000 dirs/files",
    18      'write_md5' : 'Write 1GB',
    19      'read_first_byte' : 'Time to 1st byte',
    20      'read_md5' : 'Read 1GB',
    21  }
    22  
    23  outputOrder = [
    24      'create_files',
    25      'create_files_parallel',
    26      'rm_files',
    27      'rm_files_parallel',
    28      'ls_files',
    29      'find_files',
    30      'write_md5',
    31      'read_md5',
    32      'read_first_byte',
    33  ]
    34  
    35  f = sys.argv[1]
    36  data = open(f).readlines()
    37  #print 'operation | goofys |  s3fs  | speedup'
    38  #print '----------| ------ | ------ | -------'
    39  
    40  table = [{}, {}]
    41  has_data = {}
    42  
    43  print('#operation,time')
    44  for l in data:
    45      dataset = l.strip().split('\t')
    46      for d in range(0, len(dataset)):
    47          op, num = dataset[d].split(' ')
    48          if not op in table[d]:
    49              table[d][op] = []
    50          table[d][op] += [float(num)]
    51          has_data[op] = True
    52  
    53  
    54  for c in outputOrder:
    55      if c in has_data:
    56          sys.stdout.write(op_str[c])
    57          for d in table:
    58              mean = numpy.mean(d[c])
    59              err = numpy.std(d[c])
    60              x = filter_outliers(d[c], mean, err)
    61              sys.stdout.write("\t%s\t%s\t%s" % (numpy.mean(x), numpy.min(x), numpy.max(x)))
    62          print("")
    63  
    64      # op = op_str[nums[0]]
    65  
    66      # for i in range(1, len(nums)):
    67          
    68      # x = map(lambda x: float(x), nums[1].strip().split(' '))
    69      # y = map(lambda x: float(x), nums[2].strip().split(' '))
    70      # mean_x = numpy.mean(x)
    71      # err_x = numpy.std(x)
    72      # mean_y = numpy.mean(y)
    73      # err_y = numpy.std(y)
    74      # fixed_x = fixed_y = ""
    75  
    76      # x2 = filter_outliers(x, mean_x, err_x)
    77      # y2 = filter_outliers(y, mean_y, err_y)
    78      # if x != x2:
    79      #     fixed_x = "*" * abs(len(x) - len(x2))
    80      #     mean_x = numpy.mean(x2)
    81      #     err_x = numpy.std(x2)
    82      # if y != y2:
    83      #     fixed_y = "*" * abs(len(y) - len(y2))
    84      #     mean_y = numpy.mean(y2)
    85      #     err_y = numpy.std(y2)
    86  
    87      # print "%s, %s, %s, %s", op, mean_x, mean_x - err_x, mean_x + err_x
    88      # # u_x = uncertainties.ufloat(mean_x, err_x)
    89      # # u_y = uncertainties.ufloat(mean_y, err_y)
    90      # # delta = u_y/u_x
    91      # # print "%s | %s%s | %s%s | %sx" % (op, u_x, fixed_x, u_y, fixed_y, delta)