def get_graph(node, options):
g = jsapi.QueryGraph()
#we don't use this here
# start_ts = parse_ts(options.start_ts)
#coral_fidxs['Referrer_URL'],
parsed_field_offsets = [coral_fidxs['timestamp'], coral_fidxs['HTTP_stat'],\
coral_fidxs['URL_requested'], coral_fidxs['nbytes'], \
coral_fidxs['dl_utime'], len(coral_types) ]
f = jsapi.FileRead(g, options.fname, skip_empty=True)
csvp = jsapi.CSVParse(g, coral_types)
csvp.set_cfg("discard_off_size", "true")
round = jsapi.TimeWarp(g, field=1, warp=options.warp_factor)
round.set_cfg("wait_for_catch_up", "false")
f.instantiate_on(node)
local_raw_cube = define_raw_cube(g, options.cube_name, node,
parsed_field_offsets, True)
if not options.full_url:
url_to_dom = jsapi.URLToDomain(g, field=coral_fidxs['URL_requested'])
g.chain([f, csvp, round, url_to_dom, local_raw_cube])
else:
g.chain([f, csvp, round, local_raw_cube])
return g
def src_to_badreferrer(g, data_src, node, options):
only404s = jsapi.EqualsFilter(g, field= 1, targ=404)
return g.chain([data_src, only404s, jsapi.Project(g, 5), jsapi.Project(g, 4), \
jsapi.Project(g, 2), jsapi.Project(g, 1), jsapi.URLToDomain(g, 1)])
def src_to_bad_doms(g, data_src, node, options):
return g.chain([data_src, jsapi.Project(g, 5), jsapi.Project(g, 4), \
jsapi.Project(g, 3), jsapi.URLToDomain(g, 2)])
def src_to_domain(g, data_src, node, options): url2dom = jsapi.URLToDomain(g, 2) g.chain([data_src, jsapi.Project(g, 3), url2dom]) return url2dom
def get_graph(source_nodes, root_node, options):
g= jsapi.QueryGraph()
ANALYZE = not options.load_only
LOADING = not options.analyze_only
ECHO_RESULTS = not options.no_echo
MULTIROUND = options.multiround
HASH_SAMPLE = options.hash_sample
LOCAL_THRESH = options.local_thresh
if not LOADING and not ANALYZE:
print "can't do neither load nor analysis"
sys.exit(0)
start_ts = parse_ts(options.start_ts)
central_cube = g.add_cube("global_coral_urls")
central_cube.instantiate_on(root_node)
define_cube(central_cube)
if ECHO_RESULTS:
pull_q = jsapi.TimeSubscriber(g, {}, 5000 , sort_order="-count", num_results=10)
pull_q.set_cfg("ts_field", 0)
pull_q.set_cfg("start_ts", start_ts)
pull_q.set_cfg("rollup_levels", "6,0,1") # every five seconds to match subscription. Roll up counts.
pull_q.set_cfg("simulation_rate", 1)
pull_q.set_cfg("window_offset", 6* 1000) #but trailing by a few
echo = jsapi.Echo(g)
echo.instantiate_on(root_node)
g.chain([central_cube,pull_q, echo] )
add_latency_measure(g, central_cube, root_node, tti=4, hti=5, latencylog=options.latencylog)
congest_logger = jsapi.AvgCongestLogger(g)
congest_logger.instantiate_on(root_node)
congest_logger.set_cfg("field", 3)
if MULTIROUND:
tput_merge = jsapi.MultiRoundCoord(g)
tput_merge.set_cfg("start_ts", start_ts)
tput_merge.set_cfg("window_offset", 5 * 1000)
tput_merge.set_cfg("ts_field", 0)
tput_merge.set_cfg("num_results", 10)
tput_merge.set_cfg("sort_column", "-count")
tput_merge.set_cfg("min_window_size", 5)
tput_merge.set_cfg("rollup_levels", "10,0,1") # roll up response codes
tput_merge.instantiate_on(root_node)
pull_q.set_cfg("window_offset", 10* 1000) #but trailing by a few
g.connect(tput_merge, congest_logger)
parsed_field_offsets = [coral_fidxs['timestamp'], coral_fidxs['HTTP_stat'],\
coral_fidxs['URL_requested'], len(coral_types) ]
for node, i in numbered(source_nodes, not LOADING):
if not options.full_url:
table_prefix = "local_coral_domains";
else:
table_prefix = "local_coral_urls";
table_prefix += "_"+options.warp_factor;
local_cube = g.add_cube(table_prefix+("_%d" %i))
define_cube(local_cube, parsed_field_offsets)
print "cube output dimensions:", local_cube.get_output_dimensions()
if LOADING:
f = jsapi.FileRead(g, options.fname, skip_empty=True)
csvp = jsapi.CSVParse(g, coral_types)
csvp.set_cfg("discard_off_size", "true")
round = jsapi.TimeWarp(g, field=1, warp=options.warp_factor)
if not options.full_url:
url_to_dom = jsapi.URLToDomain(g, field=coral_fidxs['URL_requested'])
g.chain( [f, csvp, round, url_to_dom, local_cube] )
else:
g.chain( [f, csvp, round, local_cube] )
f.instantiate_on(node)
else:
local_cube.set_overwrite(False)
if MULTIROUND:
pull_from_local = jsapi.MultiRoundClient(g)
else:
query_rate = 1000 if ANALYZE else 3600 * 1000
pull_from_local = jsapi.VariableCoarseningSubscriber(g, {}, query_rate)
pull_from_local.set_cfg("simulation_rate", 1)
pull_from_local.set_cfg("max_window_size", options.max_rollup)
pull_from_local.set_cfg("ts_field", 0)
pull_from_local.set_cfg("start_ts", start_ts)
pull_from_local.set_cfg("window_offset", 2000) #but trailing by a few
pull_from_local.set_cfg("sort_order", "-count")
pull_from_local.instantiate_on(node)
local_cube.instantiate_on(node)
# count_logger = jsapi.CountLogger(g, field=3)
timestamp_op= jsapi.TimestampOperator(g, "ms")
hostname_extend_op = jsapi.ExtendOperator(g, "s", ["${HOSTNAME}"]) #used as dummy hostname for latency tracker
hostname_extend_op.instantiate_on(node)
lastOp = g.chain([local_cube, pull_from_local])
if HASH_SAMPLE:
v = jsapi.VariableSampling(g, field=2, type='S')
v.set_cfg("steps", options.steps)
# print "connecting ",
lastOp = g.connect(lastOp, v)
g.add_policy( [pull_from_local, v] )
elif LOCAL_THRESH:
v = jsapi.WindowLenFilter(g)
v.set_cfg("err_field", 3)
# print "connecting ",
lastOp = g.connect(lastOp, v)
g.add_policy( [pull_from_local, v] )
g.chain( [lastOp,timestamp_op, hostname_extend_op])
#output: 0=>time, 1=>response_code, 2=> url 3=> count, 4=> timestamp at source, 5=> hostname
if MULTIROUND:
g.connect(hostname_extend_op, tput_merge)
else:
g.connect(hostname_extend_op, congest_logger)
timestamp_cube_op= jsapi.TimestampOperator(g, "ms")
timestamp_cube_op.instantiate_on(root_node)
g.chain ( [congest_logger, timestamp_cube_op, central_cube])
#input to central cube : 0=>time, 1=>response_code, 2=> url 3=> count, 4=> timestamp at source, 5=> hostname 6=> timestamp at union
if options.bw_cap:
congest_logger.set_inlink_bwcap(float(options.bw_cap))
return g
def get_graph(source_nodes, root_node, options):
ECHO_RESULTS = not options.no_echo
ANALYZE = not options.load_only
LOADING = not options.analyze_only
g= jsapi.QueryGraph()
start_ts = parse_ts(options.start_ts)
congest_logger = jsapi.AvgCongestLogger(g)
congest_logger.instantiate_on(root_node)
global_respcodes = g.add_cube("global_respcodes")
define_schema_for_cube(global_respcodes)
global_respcodes.instantiate_on(root_node)
global_ratios = g.add_cube("global_ratios")
define_schema_for_cube(global_ratios)
global_ratios.add_agg("ratio", jsapi.Cube.AggType.MIN_D, 4)
global_ratios.instantiate_on(root_node)
pull_resp = jsapi.TimeSubscriber(g, {}, 1000)
pull_resp.set_cfg("ts_field", 0)
pull_resp.set_cfg("start_ts", start_ts)
pull_resp.set_cfg("rollup_levels", "8,1,1")
pull_resp.set_cfg("simulation_rate",1)
pull_resp.set_cfg("window_offset", 5* 1000)
compute_ratio = jsapi.SeqToRatio(g, url_field = 2, total_field = 3, respcode_field = 1)
g.chain( [congest_logger, global_respcodes, pull_resp, compute_ratio, global_ratios] )
if ECHO_RESULTS:
pull_q = jsapi.TimeSubscriber(g, {}, 1000, num_results= 5, sort_order="-ratio")
pull_q.set_cfg("ts_field", 0)
pull_q.set_cfg("start_ts", start_ts)
pull_q.set_cfg("rollup_levels", "8,1,1")
pull_q.set_cfg("simulation_rate",1)
pull_q.set_cfg("window_offset", 12* 1000) #but trailing by a few
echo = jsapi.Echo(g)
echo.instantiate_on(root_node)
g.chain( [global_ratios, pull_q, echo] )
parsed_field_offsets = [coral_fidxs['timestamp'], coral_fidxs['HTTP_stat'],\
coral_fidxs['URL_requested'], len(coral_fidxs) ]
for node, i in numbered(source_nodes, False):
table_prefix = "local_coral_respcodes";
table_prefix += "_"+options.warp_factor;
local_cube = g.add_cube(table_prefix+("_%d" %i))
define_schema_for_cube(local_cube, parsed_field_offsets)
if LOADING:
f = jsapi.FileRead(g, options.fname, skip_empty=True)
csvp = jsapi.CSVParse(g, coral_types)
csvp.set_cfg("discard_off_size", "true")
round = jsapi.TimeWarp(g, field=1, warp=options.warp_factor)
round.set_cfg("wait_for_catch_up", "true")
f.instantiate_on(node)
url_to_dom = jsapi.URLToDomain(g, field=coral_fidxs['URL_requested'])
g.chain( [f, csvp, round, url_to_dom, local_cube ] )
else:
local_cube.set_overwrite(False)
query_rate = 1000 if ANALYZE else 3600 * 1000
pull_from_local = jsapi.TimeSubscriber(g, {}, query_rate)
pull_from_local.instantiate_on(node)
pull_from_local.set_cfg("simulation_rate", 1)
pull_from_local.set_cfg("ts_field", 0)
pull_from_local.set_cfg("start_ts", start_ts)
pull_from_local.set_cfg("window_offset", 2000) #but trailing by a few
local_cube.instantiate_on(node)
pull_from_local.instantiate_on(node)
g.chain( [local_cube, pull_from_local, congest_logger] )
return g
def get_graph(source_nodes, root_node, options):
g = jsapi.QueryGraph()
ANALYZE = not options.load_only
LOADING = not options.analyze_only
ECHO_RESULTS = not options.no_echo
MULTIROUND = options.multiround
HASH_SAMPLE = options.hash_sample
if not LOADING and not ANALYZE:
print "can't do neither load nor analysis"
sys.exit(0)
start_ts = parse_ts(options.start_ts)
central_cube = g.add_cube("global_coral_ua")
central_cube.instantiate_on(root_node)
define_cube(central_cube)
if ECHO_RESULTS:
pull_q = jsapi.TimeSubscriber(g, {},
5000,
sort_order="-count",
num_results=10)
pull_q.set_cfg("ts_field", 0)
pull_q.set_cfg("start_ts", start_ts)
pull_q.set_cfg("rollup_levels", "8,1")
pull_q.set_cfg("simulation_rate", 1)
pull_q.set_cfg("window_offset", 6 * 1000) #but trailing by a few
echo = jsapi.Echo(g)
echo.instantiate_on(root_node)
g.chain([central_cube, pull_q, echo])
congest_logger = jsapi.AvgCongestLogger(g)
congest_logger.instantiate_on(root_node)
congest_logger.set_cfg("field", 3)
if MULTIROUND:
tput_merge = jsapi.MultiRoundCoord(g)
tput_merge.set_cfg("start_ts", start_ts)
tput_merge.set_cfg("window_offset", 5 * 1000)
tput_merge.set_cfg("ts_field", 0)
tput_merge.set_cfg("num_results", 10)
tput_merge.set_cfg("sort_column", "-count")
tput_merge.set_cfg("min_window_size", 5)
# tput_merge.set_cfg("rollup_levels", "8,1") # roll up time
tput_merge.instantiate_on(root_node)
pull_q.set_cfg("window_offset", 10 * 1000) #but trailing by a few
g.connect(tput_merge, congest_logger)
parsed_field_offsets = [coral_fidxs['timestamp'], coral_fidxs['HTTP_stat'],\
coral_fidxs['URL_requested'], len(coral_types) ]
for node, i in numbered(source_nodes, not LOADING):
table_prefix = "local_coral_ua"
table_prefix += "_" + options.warp_factor
local_cube = g.add_cube(table_prefix + ("_%d" % i))
define_cube(local_cube, parsed_field_offsets)
print "cube output dimensions:", local_cube.get_output_dimensions()
if LOADING:
f = jsapi.FileRead(g, options.fname, skip_empty=True)
csvp = jsapi.CSVParse(g, coral_types)
csvp.set_cfg("discard_off_size", "true")
round = jsapi.TimeWarp(g, field=1, warp=options.warp_factor)
if not options.full_url:
url_to_dom = jsapi.URLToDomain(
g, field=coral_fidxs['URL_requested'])
g.chain([f, csvp, round, url_to_dom, local_cube])
else:
g.chain([f, csvp, round, local_cube])
f.instantiate_on(node)
else:
local_cube.set_overwrite(False)
if MULTIROUND:
pull_from_local = jsapi.MultiRoundClient(g)
else:
query_rate = 1000 if ANALYZE else 3600 * 1000
pull_from_local = jsapi.VariableCoarseningSubscriber(
g, {}, query_rate)
pull_from_local.set_cfg("simulation_rate", 1)
pull_from_local.set_cfg("max_window_size", options.max_rollup)
pull_from_local.set_cfg("ts_field", 0)
pull_from_local.set_cfg("start_ts", start_ts)
pull_from_local.set_cfg("window_offset",
2000) #but trailing by a few
pull_from_local.instantiate_on(node)
local_cube.instantiate_on(node)
lastOp = g.chain([local_cube, pull_from_local, congest_logger])
g.chain([congest_logger, central_cube])
return g