Automated Characterization of Health Information at Large-scale Longitudinal Evidence Systems (ACHILLES) - descriptive statistics about a OMOP CDM database
During a standard Achilles run there are dozens or even hundreds of connect - disconnect actions. A primary cause is patterns like this, where a connection is opened and closed for every SQL statement (did you know DatabaseConnector::executeSql() accepts multi-statement SQL?)
Opening and closing this many connections is inefficient, and can lead to errors (some security layers do not allow this). ideally, Achilles would open a single connection once, and close it after it's done.
Note that it is possible to open a single connection per node in a cluster, and reuse that connection across jobs, as demonstrated here.
During a standard Achilles run there are dozens or even hundreds of connect - disconnect actions. A primary cause is patterns like this, where a connection is opened and closed for every SQL statement (did you know
DatabaseConnector::executeSql()accepts multi-statement SQL?)Opening and closing this many connections is inefficient, and can lead to errors (some security layers do not allow this). ideally, Achilles would open a single connection once, and close it after it's done.
Note that it is possible to open a single connection per node in a cluster, and reuse that connection across jobs, as demonstrated here.