예제 #1
0
  def post(self):
    if self.request.get("fsm_cleanup"):
      if fsm_calculate_run_time():
        self.redirect("/subset") 
      else:
        self.response.out.write("Error calculating run time of FSM/subset") 

    if self.request.get("reset_fsm_count"):
      for c in SSFSMSimpleCounterShard.all():
        c.delete()
      self.redirect('/subset')
      return
    if self.request.get("reset_mr_count"):
      for c in SSMRSimpleCounterShard.all():
        c.delete()
      self.redirect('/subset')
      return

    if self.request.get("reset_pl_count"):
      for c in SSPLSimpleCounterShard.all():
        c.delete()
      self.redirect('/subset')
      return

    if self.request.get("compute"):
      engine = self.request.get("engine")
      dataset = self.request.get("dataset")
      user = self.request.get('user')
      data = SubSetDataSet.get_by_key_name(dataset)
      
      record = Record(engine_type=engine, 
                      dataset=dataset,
                      benchmark="subset",
                      num_entities=data.num_entries,
                      entries_per_pipe=data.entries_per_pipe,
                      user=user,
                      state="Running")
      if engine == "fsm":
        record.put()
        # reset count
        for c in SSFSMSimpleCounterShard.all():
          c.delete()

        context = {}
        context['user'] = str(user)
        context['num_entries'] = int(data.num_entries)
        fsm.startStateMachine('SubSet', [context])
        self.redirect('/subset')
      elif engine == "pipeline":
        for c in SSPLSimpleCounterShard.all():
          c.delete()
        mypipeline = SubSetPipeline(data.num_entries)
        mypipeline.start()
        record.pipeline_id = mypipeline.pipeline_id
        record.put()
        self.redirect('/subset') 
        #self.redirect(mypipeline.base_path + "/status?root=" + mypipeline.pipeline_id)
      elif engine == "mr":
        for c in SSMRSimpleCounterShard.all():
          c.delete()
        # Why 1k each per shard or less? is this ideal?
        if data.num_entries > 1000: shards = data.num_entries/1000
        else: shards = 1

        kind = get_class(data.num_entries)
        mapreduce_id = control.start_map(
          name="Wordcount with just mappers",
          handler_spec="subset.mr.subset_mapper",
          reader_spec="mapreduce.input_readers.DatastoreInputReader",
          mapper_parameters={
              "entity_kind": "data.subset."+kind,
              "processing_rate": 500
          },
          mapreduce_parameters={model.MapreduceSpec.PARAM_DONE_CALLBACK:
                     '/subset/mr/callback'},
          shard_count=shards,
          queue_name="default",
        )

        record.mr_id = mapreduce_id
        record.put()
        self.redirect('/subset')
예제 #2
0
  def post(self):
    """ Generate data sets here """
    if self.request.get("fsm_cleanup"):
      if fsm_calculate_run_time():
        self.redirect('/grep')
      else:
        self.response.out.write("Error calculating fsm/grep")
      return 
    if self.request.get("compute"):
      engine = self.request.get("engine")
      dataset = self.request.get("dataset")
      user = self.request.get('user')
      needle = self.request.get('needle')    
      data = GrepDataSet.get_by_key_name(dataset)
      record = Record(engine_type=engine, 
                      dataset=dataset,
                      benchmark="grep",
                      num_entities=data.num_entries,
                      #shard_count=data.num_pipelines,
                      entries_per_pipe=data.entries_per_pipe,
                      user=user,
                      char_per_word=data.char_per_word,
                      state="Running")
      if engine == "fsm":
        record.put()
        context = {}
        context['user'] = str(user)
        context['num_entries'] = int(data.num_entries)
        context['needle'] = needle
        fsm.startStateMachine('Grep', [context])
        self.redirect('/grep')
      elif engine == "pipeline":
        mypipeline = GrepPipelineLoop(data.num_entries, needle)
        mypipeline.start()
        record.pipeline_id = mypipeline.pipeline_id
        record.put()
        self.redirect('/grep')
        #self.redirect(mypipeline.base_path + "/status?root=" + mypipeline.pipeline_id)
        return
      elif engine == "mr":
        # Why 1k each per shard or less? is this ideal?
        if data.num_entries > 1000:
          shards = data.num_entries/1000
          shards = min(256, shards) 
        else: shards = 1

        kind = getKindString(data.num_entries)
        mapreduce_id = control.start_map(
            name="Grep",
            handler_spec="grep.mr.grep_mapper",
            reader_spec="mapreduce.input_readers.DatastoreInputReader",
            mapper_parameters={
                "entity_kind": "data.grep."+kind,
                "processing_rate": 500,
                "needle":needle,
            },
            mapreduce_parameters={model.MapreduceSpec.PARAM_DONE_CALLBACK:
                       '/grep/mr/callback'},
            shard_count=shards,
            queue_name="default",
          )

        record.mr_id = mapreduce_id
        record.put()
        self.redirect('/grep')