Access the metric values through macros

To simplify the access to some specific values defined inside the SeD , you can use these macros:

-

TOTALTIME(SeD)

-

COMMTIME(SeD)

-

TCOMP(SeD)

-

TIMESINCELASTSOLVE(SeD)

-

COMMPROXIMITY(SeD)

-

TRANSFEREFFORT(SeD)

-

FREECPU(SeD)

-

FREEMEM(SeD)

-

NBCPU(SeD)

-

CPUSPEED(SeD, idx)

-

TOTALMEM(SeD)

-

AVGFREEMEM(SeD)

-

AVGFREECPU(SeD)

-

BOGOMIPS(SeD, idx)

-

CACHECPU(SeD, idx)

-

TOTALSIZEDISK(SeD)

-

FREESIZEDISK(SeD)

-

DISKACCESSREAD(SeD)

-

DISKACCESSWRITE(SeD)

-

USERDEFINED(SeD, idx)

The macros taking two arguments need an index to choose which CPU measurement is needed. Two extra macros are defined:

• HOSTNAME(server): The hostname of the SeD .
• SED_REF(server): A CORBA reference to the SeD .

Here is an example of an aggregate function using these macros:

SORTFUN(compBogo, BOGOMIPS)

int MyScheduler::aggregate(corba_response_t* aggrResp, size_t max_srv,
                           const size_t nb_responses,
                           const corba_response_t* responses)
{
  ServerList candidates = CORBA_to_STL(responses, nb_responses);
  ServerList chosen;
  ServerList::iterator it;

  for (it=candidates.begin(); it!=candidates.end(); ++it)
    if (NBCPU(*it)>=2) chosen.push_back(*it);
  SORT(chosen, compBogo);

  STL_to_CORBA(chosen, aggrResp);
  return 0;
}

This aggregation method first selects only the SeD which have more than 1 CPU and sorts them according to their number of Bogomips.