use time.clock() instead of time.time()

[c-nichols]

time.time() will return the server's best guess at the current wall clock time. Wall clock time can be affected by NTP updates, result in artificially high or low (even negative) runtimes for requests.

time.clock() uses the processor's internal counters to measure time since last reboot. It will never "jump" like time.time() will, and so is the preferred method for profiling runtimes.