Google claims that GAE/J creates new application instances nearly instantly (and using some optimised re-allocation techniques if free instances are available) with incoming requests to the web front-end of an application. However, GAE/J penalises applications that have higher “application latency”, which Google defines as the turnaround time for an application to process a user request. The higher this latency, the more conservative GAE/J is in terms of allocating new resources. GAE/J treats a value ≥ 1,000ms as a bad application latency, which is not unusual for a servlet-based application performing complex cryptographic operations.