def site_size_parser(sql_results, globals=globals(), **kw):
"""
Take in CPU hours information (# of wall hours per timespan per site)
and convert it into the "size" of the site as a function of time.
"""
#print "in site size parser"
results, md = results_parser(sql_results, globals=globals, **kw)
span = kw['span']
hours = span / 3600.
sites = results.keys()
new_results = {}
all_intervals = sets.Set()
for site in sites:
intervals = results[site].keys()
all_intervals.union_update(intervals)
#print len(all_intervals)
all_intervals = list(all_intervals)
all_intervals.sort()
for site in sites:
new_results[site] = {}
mymax = NMax(2)
for start in all_intervals:
avg_cpus = results[site].get(start, 0) / hours
mymax.add_datum(avg_cpus)
new_results[site][start] = mymax.get_max()
return new_results, md
def site_size_parser(sql_results, globals=globals(), **kw):
"""
Take in CPU hours information (# of wall hours per timespan per site)
and convert it into the "size" of the site as a function of time.
"""
#print "in site size parser"
results, md = results_parser(sql_results, globals=globals, **kw)
span = kw['span']
hours = span / 3600.
sites = results.keys()
new_results = {}
all_intervals = sets.Set()
for site in sites:
intervals = results[site].keys()
all_intervals.union_update(intervals)
#print len(all_intervals)
all_intervals = list(all_intervals)
all_intervals.sort()
for site in sites:
new_results[site] = {}
mymax = NMax(2)
for start in all_intervals:
avg_cpus = results[site].get(start, 0) / hours
mymax.add_datum(avg_cpus)
new_results[site][start] = mymax.get_max()
return new_results, md
def non_physics_filter(sql_results, globals=globals(), **kw):
"""
Removes results for Physics VOs.
"""
results, md = results_parser(sql_results, globals=globals, **kw)
hep_vos = physics_classifier(results.keys(), globals=globals)
#print "HEP VOs"
#print "\n".join(hep_vos)
filtered_results = {}
for pivot, group in results.items():
if pivot not in hep_vos:
filtered_results[pivot] = group
return filtered_results, md
def non_physics_filter(sql_results, globals=globals(), **kw):
"""
Removes results for Physics VOs.
"""
results, md = results_parser(sql_results, globals=globals, **kw)
hep_vos = physics_classifier(results.keys(), globals=globals)
#print "HEP VOs"
#print "\n".join(hep_vos)
filtered_results = {}
for pivot, group in results.items():
if pivot not in hep_vos:
filtered_results[pivot] = group
return filtered_results, md
def science_classifier(sql_results, globals=globals(), default="Other", **kw):
"""
Take in some VO-based metric and convert it to a field of science-based
metric. Uses the fact that the field of science is recorded by OIM.
"""
results, md = results_parser(sql_results, globals=globals, **kw)
fields_of_science, _ = globals['RSVQueries'].field_of_science()
fields_of_science += addl_fields_of_science
gratia_vos = results.keys()
#print fields_of_science
oim_vos = [i[0] for i in fields_of_science]
oim_to_gratia, gratia_to_oim = OIM_to_gratia_mapper(oim_vos, gratia_vos)
vo_to_science = {}
for oim_vo, science_field in fields_of_science:
current_science = vo_to_science.get(oim_vo, '')
precedence_cur = precedence.get(current_science, 99)
precedence_new = precedence.get(science_field)
#print oim_vo, science_field, precedence_new, current_science, precedence_cur
if precedence_new < precedence_cur:
vo_to_science[oim_vo] = science_field
#print "Gratia VO to Science"
#for vo, science in vo_to_science.items():
# print vo, science
filtered_results = {}
for pivot, groups in results.items():
if pivot in gratia_to_oim and gratia_to_oim[pivot] in vo_to_science:
new_pivot = vo_to_science[gratia_to_oim[pivot]]
else:
#print "Unclassified VO:", pivot
new_pivot = default
if new_pivot == 'HEP':
new_pivot = 'High Energy Physics'
if new_pivot not in filtered_results:
filtered_results[new_pivot] = groups
else:
for group, val in groups.items():
cur = filtered_results[new_pivot].get(group, 0)
filtered_results[new_pivot][group] = cur + val
if 'Physics' in filtered_results:
filtered_results['non-HEP Physics'] = filtered_results['Physics']
del filtered_results['Physics']
return filtered_results, md
def science_classifier(sql_results, globals=globals(), default="Other", **kw):
"""
Take in some VO-based metric and convert it to a field of science-based
metric. Uses the fact that the field of science is recorded by OIM.
"""
results, md = results_parser(sql_results, globals=globals, **kw)
fields_of_science, _ = globals['RSVQueries'].field_of_science()
fields_of_science += addl_fields_of_science
gratia_vos = results.keys()
#print fields_of_science
oim_vos = [i[0] for i in fields_of_science]
oim_to_gratia, gratia_to_oim = OIM_to_gratia_mapper(oim_vos, gratia_vos)
vo_to_science = {}
for oim_vo, science_field in fields_of_science:
current_science = vo_to_science.get(oim_vo, '')
precedence_cur = precedence.get(current_science, 99)
precedence_new = precedence.get(science_field)
#print oim_vo, science_field, precedence_new, current_science, precedence_cur
if precedence_new < precedence_cur:
vo_to_science[oim_vo] = science_field
#print "Gratia VO to Science"
#for vo, science in vo_to_science.items():
# print vo, science
filtered_results = {}
for pivot, groups in results.items():
if pivot in gratia_to_oim and gratia_to_oim[pivot] in vo_to_science:
new_pivot = vo_to_science[gratia_to_oim[pivot]]
else:
#print "Unclassified VO:", pivot
new_pivot = default
if new_pivot == 'HEP':
new_pivot = 'High Energy Physics'
if new_pivot not in filtered_results:
filtered_results[new_pivot] = groups
else:
for group, val in groups.items():
cur = filtered_results[new_pivot].get(group, 0)
filtered_results[new_pivot][group] = cur + val
if 'Physics' in filtered_results:
filtered_results['non-HEP Physics'] = filtered_results['Physics']
del filtered_results['Physics']
return filtered_results, md
def osg_site_size(sql_results, globals=globals(), **kw):
"""
Calculate the OSG's size in terms of utilized CPUs, accessible CPUs, and
total CPUs. Break down these statistics by site.
"""
USED = 'Max Used'
UNACCESSIBLE = 'In OSG, but never used'
if 'normalize' in kw and kw['normalize'].lower().find('t') >= 0:
normalize = True
else:
normalize = False
utilized_results, md = results_parser(sql_results, globals=globals, **kw)
accessible_results, _ = globals['GratiaBarQueries'].osg_avail_size(
span=7 * 86400, starttime=time.time() - 7 * 86400 * 52)
total_results, _ = globals['GIPQueries'].gip_site_size(
span=7 * 86400, starttime=time.time() - 7 * 86400 * 52, max_size=20000)
ksi2k_results, _ = globals['GIPQueries'].subcluster_score_ts()
ksi2k_results2, _ = globals['GIPQueries'].subcluster_score_ts2()
ksi2k_results2 = ksi2k_results2['Nebraska']
sites = utilized_results.keys()
new_results = {}
all_intervals = sets.Set()
for site in sites:
intervals = utilized_results[site].keys()
all_intervals.union_update(intervals)
all_intervals = list(all_intervals)
all_intervals.sort()
total_utilized_results = {}
total_accessible_results = {}
total_total_results = {}
final_results = {USED: {}, ACCESSIBLE: {}, UNACCESSIBLE: {}}
may_1 = time.mktime((2008, 05, 01, 0, 0, 0, 0, 0, 0))
avg_ksi2k_results = {}
ksi2k_min = min(1.7, ksi2k_results2.values())
ksi2k_max = ksi2k_min
for interval in all_intervals:
ksi2k_max = max(ksi2k_results2.get(interval, ksi2k_min), ksi2k_max)
avg_ksi2k_results[interval] = ksi2k_max
prev_interval = 0
for interval in all_intervals:
# Process accessible numbers
current_acc = 0
for site, vals in accessible_results.items():
if site not in ksi2k_results:
ksi2k = avg_ksi2k_results[interval]
elif interval not in ksi2k_results[site]:
ksi2k = min(min(ksi2k_results[site].values()),
avg_ksi2k_results[interval])
else:
ksi2k = ksi2k_results[site][interval]
if normalize:
current_acc = vals.get(interval, 0) * ksi2k
else:
current_acc = vals.get(interval, 0)
prev_acc = total_accessible_results.setdefault(site, 0)
total_accessible_results[site] = max(prev_acc, current_acc)
# Process total size numbers
cumulative3 = 0
for site, vals in total_results.items():
if site not in ksi2k_results:
ksi2k = avg_ksi2k_results[interval]
elif interval not in ksi2k_results[site]:
ksi2k = min(min(ksi2k_results[site].values()),
avg_ksi2k_results[interval])
else:
ksi2k = ksi2k_results[site][interval]
if normalize:
curr_total = vals.get(interval, 0) * ksi2k
else:
curr_total = vals.get(interval, 0)
prev_total = total_total_results.setdefault(site, 0)
total_total_results[site] = max(prev_total, curr_total)
if interval < may_1:
continue
for site in sites:
# Update the final results
final_results[USED][site] = total_accessible_results.get(site, 0)
final_results[UNACCESSIBLE][site] = max(total_total_results.get( \
site, 0) - total_accessible_results.get(site, 0), 0)
return final_results, md
def osg_size(sql_results, globals=globals(), **kw):
"""
Calculate the OSG's size in terms of utilized CPUs, accessible CPUs, and
total CPUs..
"""
if 'normalize' in kw and kw['normalize'].lower().find('t') >= 0:
normalize = True
else:
normalize = False
utilized_results, md = results_parser(sql_results, globals=globals, **kw)
accessible_results, _ = globals['GratiaBarQueries'].osg_avail_size(
span=7 * 86400, starttime=time.time() - 7 * 86400 * 52)
total_results, _ = globals['GIPQueries'].gip_site_size(
span=7 * 86400, starttime=time.time() - 7 * 86400 * 52, max_size=20000)
ksi2k_results, _ = globals['GIPQueries'].subcluster_score_ts()
ksi2k_results2, _ = globals['GIPQueries'].subcluster_score_ts2()
ksi2k_results2 = ksi2k_results2['Nebraska']
sites = utilized_results.keys()
new_results = {}
all_intervals = sets.Set()
for site in sites:
intervals = utilized_results[site].keys()
all_intervals.union_update(intervals)
all_intervals = list(all_intervals)
all_intervals.sort()
total_utilized_results = {}
total_accessible_results = {}
total_total_results = {}
final_results = {
'Used': {},
'Accessible, but not Used': {},
'In OSG, but not Accessible': {}
}
may_1 = time.mktime((2008, 05, 01, 0, 0, 0, 0, 0, 0))
avg_ksi2k_results = {}
ksi2k_min = min(1.7, ksi2k_results2.values())
ksi2k_max = ksi2k_min
for interval in all_intervals:
ksi2k_max = max(ksi2k_results2.get(interval, ksi2k_min), ksi2k_max)
avg_ksi2k_results[interval] = ksi2k_max
prev_interval = 0
for interval in all_intervals:
cumulative = 0
for site, vals in utilized_results.items():
if site not in ksi2k_results:
ksi2k = avg_ksi2k_results[interval]
elif interval not in ksi2k_results[site]:
ksi2k = min(min(ksi2k_results[site].values()),
avg_ksi2k_results[interval])
else:
ksi2k = ksi2k_results[site][interval]
if normalize:
cumulative += vals.get(interval, 0) * ksi2k
else:
cumulative += vals.get(interval, 0)
total_utilized_results[interval] = cumulative
cumulative2 = 0
for site, vals in accessible_results.items():
if site not in ksi2k_results:
ksi2k = avg_ksi2k_results[interval]
elif interval not in ksi2k_results[site]:
ksi2k = min(min(ksi2k_results[site].values()),
avg_ksi2k_results[interval])
else:
ksi2k = ksi2k_results[site][interval]
if normalize:
cumulative2 += vals.get(interval, 0) * ksi2k
else:
cumulative2 += vals.get(interval, 0)
total_accessible_results[interval] = cumulative2
cumulative3 = 0
for site, vals in total_results.items():
if site not in ksi2k_results:
ksi2k = avg_ksi2k_results[interval]
elif interval not in ksi2k_results[site]:
ksi2k = min(min(ksi2k_results[site].values()),
avg_ksi2k_results[interval])
else:
ksi2k = ksi2k_results[site][interval]
if normalize:
cumulative3 += vals.get(interval, 0) * ksi2k
else:
cumulative3 += vals.get(interval, 0)
total_total_results[interval] = cumulative3
if interval < may_1:
continue
final_results[USED][interval] = cumulative
final_results[ACCESSIBLE][interval] = max(cumulative2 -\
cumulative, 0)
final_results[UNACCESSIBLE][interval] = max(cumulative3\
- cumulative2, 0)
# Make sure numbers never go down.
# This should be true because all the numbers should be cumulative,
# but we're just being paranoid here.
#for pivot in [ACCESSIBLE, UNACCESSIBLE]:
# if prev_interval in final_results[pivot] and final_results[pivot]\
# [prev_interval] > final_results[pivot][interval]:
# final_results[pivot][interval] = final_results[pivot]\
# [prev_interval]
#prev_interval = interval
return final_results, md
def osg_site_size(sql_results, globals=globals(), **kw):
"""
Calculate the OSG's size in terms of utilized CPUs, accessible CPUs, and
total CPUs. Break down these statistics by site.
"""
USED = 'Max Used'
UNACCESSIBLE = 'In OSG, but never used'
if 'normalize' in kw and kw['normalize'].lower().find('t') >= 0:
normalize = True
else:
normalize = False
utilized_results, md = results_parser(sql_results, globals=globals, **kw)
accessible_results, _ = globals['GratiaBarQueries'].osg_avail_size(span=7*86400,
starttime=time.time()-7*86400*52)
total_results, _ = globals['GIPQueries'].gip_site_size(span=7*86400,
starttime=time.time()-7*86400*52, max_size=20000)
ksi2k_results, _ = globals['GIPQueries'].subcluster_score_ts()
ksi2k_results2, _ = globals['GIPQueries'].subcluster_score_ts2()
ksi2k_results2 = ksi2k_results2['Nebraska']
sites = utilized_results.keys()
new_results = {}
all_intervals = sets.Set()
for site in sites:
intervals = utilized_results[site].keys()
all_intervals.union_update(intervals)
all_intervals = list(all_intervals)
all_intervals.sort()
total_utilized_results = {}
total_accessible_results = {}
total_total_results = {}
final_results = {USED: {}, ACCESSIBLE: {}, UNACCESSIBLE: {}}
may_1 = time.mktime((2008, 05, 01, 0, 0, 0, 0, 0, 0))
avg_ksi2k_results = {}
ksi2k_min = min(1.7, ksi2k_results2.values())
ksi2k_max = ksi2k_min
for interval in all_intervals:
ksi2k_max = max(ksi2k_results2.get(interval, ksi2k_min), ksi2k_max)
avg_ksi2k_results[interval] = ksi2k_max
prev_interval = 0
for interval in all_intervals:
# Process accessible numbers
current_acc = 0
for site, vals in accessible_results.items():
if site not in ksi2k_results:
ksi2k = avg_ksi2k_results[interval]
elif interval not in ksi2k_results[site]:
ksi2k = min(min(ksi2k_results[site].values()),
avg_ksi2k_results[interval])
else:
ksi2k = ksi2k_results[site][interval]
if normalize:
current_acc = vals.get(interval, 0) * ksi2k
else:
current_acc = vals.get(interval, 0)
prev_acc = total_accessible_results.setdefault(site, 0)
total_accessible_results[site] = max(prev_acc, current_acc)
# Process total size numbers
cumulative3 = 0
for site, vals in total_results.items():
if site not in ksi2k_results:
ksi2k = avg_ksi2k_results[interval]
elif interval not in ksi2k_results[site]:
ksi2k = min(min(ksi2k_results[site].values()),
avg_ksi2k_results[interval])
else:
ksi2k = ksi2k_results[site][interval]
if normalize:
curr_total = vals.get(interval, 0) * ksi2k
else:
curr_total = vals.get(interval, 0)
prev_total = total_total_results.setdefault(site, 0)
total_total_results[site] = max(prev_total, curr_total)
if interval < may_1:
continue
for site in sites:
# Update the final results
final_results[USED][site] = total_accessible_results.get(site, 0)
final_results[UNACCESSIBLE][site] = max(total_total_results.get( \
site, 0) - total_accessible_results.get(site, 0), 0)
return final_results, md
def osg_size(sql_results, globals=globals(), **kw):
"""
Calculate the OSG's size in terms of utilized CPUs, accessible CPUs, and
total CPUs..
"""
if 'normalize' in kw and kw['normalize'].lower().find('t') >= 0:
normalize = True
else:
normalize = False
utilized_results, md = results_parser(sql_results, globals=globals, **kw)
accessible_results, _ = globals['GratiaBarQueries'].osg_avail_size(span=7*86400,
starttime=time.time()-7*86400*52)
total_results, _ = globals['GIPQueries'].gip_site_size(span=7*86400,
starttime=time.time()-7*86400*52, max_size=20000)
ksi2k_results, _ = globals['GIPQueries'].subcluster_score_ts()
ksi2k_results2, _ = globals['GIPQueries'].subcluster_score_ts2()
ksi2k_results2 = ksi2k_results2['Nebraska']
sites = utilized_results.keys()
new_results = {}
all_intervals = sets.Set()
for site in sites:
intervals = utilized_results[site].keys()
all_intervals.union_update(intervals)
all_intervals = list(all_intervals)
all_intervals.sort()
total_utilized_results = {}
total_accessible_results = {}
total_total_results = {}
final_results = {'Used': {}, 'Accessible, but not Used': {},
'In OSG, but not Accessible': {}}
may_1 = time.mktime((2008, 05, 01, 0, 0, 0, 0, 0, 0))
avg_ksi2k_results = {}
ksi2k_min = min(1.7, ksi2k_results2.values())
ksi2k_max = ksi2k_min
for interval in all_intervals:
ksi2k_max = max(ksi2k_results2.get(interval, ksi2k_min), ksi2k_max)
avg_ksi2k_results[interval] = ksi2k_max
prev_interval = 0
for interval in all_intervals:
cumulative = 0
for site, vals in utilized_results.items():
if site not in ksi2k_results:
ksi2k = avg_ksi2k_results[interval]
elif interval not in ksi2k_results[site]:
ksi2k = min(min(ksi2k_results[site].values()),
avg_ksi2k_results[interval])
else:
ksi2k = ksi2k_results[site][interval]
if normalize:
cumulative += vals.get(interval, 0) * ksi2k
else:
cumulative += vals.get(interval, 0)
total_utilized_results[interval] = cumulative
cumulative2 = 0
for site, vals in accessible_results.items():
if site not in ksi2k_results:
ksi2k = avg_ksi2k_results[interval]
elif interval not in ksi2k_results[site]:
ksi2k = min(min(ksi2k_results[site].values()),
avg_ksi2k_results[interval])
else:
ksi2k = ksi2k_results[site][interval]
if normalize:
cumulative2 += vals.get(interval, 0) * ksi2k
else:
cumulative2 += vals.get(interval, 0)
total_accessible_results[interval] = cumulative2
cumulative3 = 0
for site, vals in total_results.items():
if site not in ksi2k_results:
ksi2k = avg_ksi2k_results[interval]
elif interval not in ksi2k_results[site]:
ksi2k = min(min(ksi2k_results[site].values()),
avg_ksi2k_results[interval])
else:
ksi2k = ksi2k_results[site][interval]
if normalize:
cumulative3 += vals.get(interval, 0) * ksi2k
else:
cumulative3 += vals.get(interval, 0)
total_total_results[interval] = cumulative3
if interval < may_1:
continue
final_results[USED][interval] = cumulative
final_results[ACCESSIBLE][interval] = max(cumulative2 -\
cumulative, 0)
final_results[UNACCESSIBLE][interval] = max(cumulative3\
- cumulative2, 0)
# Make sure numbers never go down.
# This should be true because all the numbers should be cumulative,
# but we're just being paranoid here.
#for pivot in [ACCESSIBLE, UNACCESSIBLE]:
# if prev_interval in final_results[pivot] and final_results[pivot]\
# [prev_interval] > final_results[pivot][interval]:
# final_results[pivot][interval] = final_results[pivot]\
# [prev_interval]
#prev_interval = interval
return final_results, md
