github.com/sagansystems/goofys-app@v0.19.1-0.20180410053237-b2302fdf5af9/bench/bench_format.py (about)

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