TODO

* Extend benchmarks to run correctly also in a virtualized environment:

  . add a "guest mode" to the benchmarks, for which the following
    functionalities are provided

  . choice of the host I/O scheduler

  . possibility to start a background workload on the host
    before starting the benchmark in the guest

    . the script executed in the host may send sowmehow a signal (for
      example by creating a file in some shared folder, or by executing
      a kill with the proper signal in the guest, ...) when the workload
      has settled in the host, i.e., after the usual initial sleep

	. after invoking the host script, the script executed in the guest
	  blocks until the above signal is received
	
	. at the end of the bechmark, the guest script sends somehow a
	  signal to the host script (for example by executing a kill
	  with the right signal in the host), which shuts down the
	  workload (hence the host script must run the workload
	  indefinitely, till it receives the stop signal)

  . sync&drop or just drop caches also on the host when appropriate

  . make sure that also the host-related actions are logged on stdout,
    so that the guest log contains all the information one may need to
    check whether the benchmark has been correctly executed

  . automatically add an extra tag just after the scheduler name in
    the name of the output stat file, for example with the following
    format, where <SCHED> is the name of the host scheduler in
    uppercase (to help back-end scripts not get confused with the
    guest scheduler):

      . no_host_wl-<SCHED>			If no workload is run on the host

      . host_<n>r[<m>w][_<type>]-<SCHED>- 	If n readers and m writers
      						of some type are run in the
      						host, in particular <m>w- is
						added only if m>0, and 
						_<type> is added only if it is
						not seq

  . write the above tag and not just the scheduler name in the output stat file

* Using the above new functionalities of the benchamrk scripts, define
  a new run_all_test script for virtualized environments

  . for each benchmark, run consecutively the benchmark script for all
    possible guest and host configurations, and, for each repetition,
    put all output stat files in the same folder

  . possibily merge this run_all_tests script with the existing
    run_all_tests script