This system procedure will cause every HStoreSite in the cluster to recompute the probabilities of their transaction prediction models. If the the save_to_file input parameter is set to true, then the procedure will write out the updated Markov models to a random file in the HStoreSite’s global.temp_dir.
You can have the BenchmarkController invoke this procedure automatically after the warm-up period of a benchmark by passing in the markov.recompute_warmup parameter, or after the entire benchmark run by passing in the the markov.recompute_end parameter. This will automatically retrieve the written files and combine them into a single MarkovGraphsContainer file that can then be loaded into the system.
||Whether the procedure should write out the update Markov models to a temporary file on disk.
||The id of the site that recomputed its models.
||The id of the partition that recomputed its models.
||The path to file that the HStoreSite saved its updated Markov models for this partition. If save_to_file is false, then this field will be null.
||Whether or not the Markov models are global (i.e., they are not partitioned with a decision tree). (1 = TRUE, 0 = FALSE)
Client client = ClientFactory.createClient();
client = client.createConnection(hostname, port);
ClientResponse cresponse = client.callProcedure("@MarkovUpdate", 1);
hstore> EXEC @MarkovUpdate 1