def shuffle(self, input_source):
results = []
#debug_print("[Wide-RDD] {0} InputSource is {1}".format(self.id, input_source))
for source in input_source:
# debug_print("[Shuffle] {0} Shuffling from source {1}".format(self.id, source))
if not isinstance(source, list):
source=[source]
for p in source:
result=None
while result is None:
task_node_table=p["task_node_table"]
worker_address= task_node_table["{0}_{1}".format(p['job_id'],p['task_id'])]["address"]
client = get_client(worker_address)
# debug_print("[Shuffle] {0} get a None from {1}, at Part {2}, retrying".format(self.id, p['task_id'],p['partition_id']))
result=execute_command(client, client.get_rdd_result,
p['job_id'],
p['task_id'],
p['partition_id'])
# debug_print("[Shuffle] {0} get a result={1} from {2}, at Part {3}".format(self.id, result, p['task_id'],p['partition_id']))
results += result
return results
'num_partition_GBK': 2,
'split_size': 128,
}
t = rdd.TextFile(self.filename)
m = rdd.Map(t, (lambda urls: parseNeighbors(urls)))
links = rdd.GroupByKey(m)
ranks = rdd.Map(links, lambda url_neighbors: (url_neighbors[0], 1.0))
for iteration in range(5):
joins = rdd.Join([links, ranks])
contribs = rdd.FlatMap(
joins, lambda url_urls_rank: computeContribs(
url_urls_rank[1][0], url_urls_rank[1][1]))
rbk = rdd.ReduceByKey(contribs, lambda a, b: a + b)
ranks = rdd.MapValue(rbk, lambda rank: rank * 0.85 + 0.15)
ranks.collect(driver)
if __name__ == '__main__':
master_address = sys.argv[1]
self_address = sys.argv[2]
filepath = sys.argv[3]
page_rank_client = PageRankClient(filepath)
# page_rank_client = PageRankClient(sys.argv[1])
client = get_client(master_address)
execute_command(client, client.get_job, pickle_object(page_rank_client),
self_address)
print "[Client]Job Submited...."
page_rank_client.start_server(self_address)
RDD._config = {'num_partition_RBK': 2,
'num_partition_GBK': 2,
'split_size': 128,
}
t = rdd.TextFile(self.filename)
m = rdd.Map(t, (lambda urls: parseNeighbors(urls)))
links = rdd.GroupByKey(m)
ranks = rdd.Map(links, lambda url_neighbors: (url_neighbors[0], 1.0))
for iteration in range(5):
joins = rdd.Join([links, ranks])
contribs = rdd.FlatMap(joins,
lambda url_urls_rank: computeContribs(url_urls_rank[1][0], url_urls_rank[1][1]))
rbk = rdd.ReduceByKey(contribs, lambda a, b: a + b)
ranks = rdd.MapValue(rbk, lambda rank: rank * 0.85 + 0.15)
ranks.collect(driver)
if __name__ == '__main__':
master_address = sys.argv[1]
self_address = sys.argv[2]
filepath = sys.argv[3]
page_rank_client = PageRankClient(filepath)
# page_rank_client = PageRankClient(sys.argv[1])
client = get_client(master_address)
execute_command(client, client.get_job, pickle_object(page_rank_client), self_address)
print "[Client]Job Submited...."
page_rank_client.start_server(self_address)
