The classical calculator, as all the calculators except the event based one, is based on a custom distribution mechanism built on top of kombu, which is implementing a task pool. Such approach has a lot of know disadvantages:
1. it is not the standard celery distribution mechanism
2. it is cumbersome to use: in particular it is very easy to forget to update the progress counter correctly, thus having a calculator that hangs (this happened several time when developing new calculators)
3. it is terrible with error reporting: the errors are logged and a supervisor is needed, just to parse the log files and to kill the computation is an error occurs; in Jenkins and the engine-server the supervisor is missing, so an error in a task hangs the whole computation
4. the distribution is tightly coupled with a custom progress mechanism which is complex and requires Redis as a dependency: so, it is hard to port on Windows, an issue in view of the engine-lite project
5. the distribution does not limit the number of tasks generated, so it is pretty easy to use a block_size too small, generate a lot of tasks and having a calculation that does nothing except passing messages around with rabbitmq.
The event based calculator has a distribution mechanism (the "parallelize" mechanism) with none of these issues, so it makes sense to use it for all calculators.
https:/