Python - Are Threads or Processes more appropriate for disk I/O-bound operations?

iotest.py

from threading import Thread
from multiprocessing import Process
from os import remove

from time import time

class IOTest:
    max_workers = 10
    max_lines = 10000000
    
    max_attempts = 10
    
    file_location = '/data/db'
    
    worker_model = None
    
    def __init__(self):
        self.name = self.__class__.__name__
    
    def write(self, pid):
        p = self.__class__.__name__ + str(pid)
        filename = '%s/%s.txt' % (self.file_location, p)
        with open(filename, 'w') as f:
            for x in xrange(self.max_lines):
                f.write('Nothing signed "THE MGT." would ever be challenged; the Midget could always pass himself off as the Management.\n')
        remove(filename)
                
    def run(self):
        results = []
        print 'Starting', self.name, '...'
        for attempt in xrange(self.max_attempts):
            # Create N workers
            workers = []
            for y in xrange(self.max_workers):
                workers.append(self.worker_model(target=self.write, args=(y,)))
                
            start_time = time()
            [t.start() for t in workers]
            [t.join() for t in workers]
            elapsed = time() - start_time
            results.append(elapsed)
            print '%s attempt %i of %i took %s seconds to write %i lines to %i files.' % (self.name, attempt+1, self.max_attempts, elapsed, self.max_lines, self.max_workers)
            
        print '-' * 80
        print self.name, 'results:', [round(x,2) for x in results]
        print 'avg:', sum(results) / len(results)
        print '-' * 80
        
        return sum(results) / len(results)
    
class Multithreaded(IOTest):
    worker_model = Thread
    
class Multiprocessed(IOTest):
    worker_model = Process
    
thread_avg = Multithreaded().run()
process_avg = Multiprocessed().run()

print '-' * 80
if thread_avg < process_avg:
    print 'Threading is faster.'
else:
    print 'Multiprocessing is faster.'