Model engines & cloud computing
To successfully run a model using OpenAgua, there must be a working connection between OpenAgua and a model engine. A model engine is a generalized executable modeling application that is designed to 1) connect with and read data from the OpenAgua database, 2) perform some action using the data (most likely a simulation or optimization of the system), and 3) save results back to OpenAgua.
There is no hard constraint on what action is performed within the model engine, so this setup actually affords a significant amount of flexibility in terms of customization of the model, from the trivial (e.g., do nothing) to the complex (e.g., simulate a coupled human-water system using multiple sub-models in various physical and management domains).
Conceptual overview
When running a model, a user initiates a set of actions that results in the model engine starting. The model engine may be located on one or more computers connected to the internet. Generally, there are three modes that OpenAgua can use to communicate with a model engine.
Directly on the host computer. In this case, the model engine is located on the same computer as the OpenAgua server. OpenAgua sends the command to run the model engine directly (using Python's Popen constructor).
Directly via SSH on a remote computer: In this case, the model engine is located on a computer different than the main OpenAgua server. OpenAgua sends the command to run the model using SSH (secure shell). SSH is a way of logging directly into a computer remotely to execute commands on that computer.
Indirectly via a task [scheduler](https://en.wikipedia.org/wiki/Scheduling_(computing)): In this case, the model engine may be on one or more computers, and may include the OpenAgua server itself. Running the model involves two steps behind-the-scenes. First, OpenAgua sends the command to a scheduler (specifically, a RabbitMQ task queue). Second, one of the model engines, which are setup as workers to continually wait for new tasks, takes the task out of the queue and performs the model computations.
The practical implications of these are that model engine setup requires configuration in two places: in OpenAgua and in the model engine itself. Both of these are described here.
Cloud computing
*Cloud computing* is a widely used, and at times ambiguous term. Here, *cloud computing* refers to the ability to use any computer connected to the internet for computationally challenging tasks, including modeling. This is achieved using the third method above, i.e. via a task scheduler. This allows any cloud-connected computer to be used for the computational part of modeling. Examples include:
Commercial Infrastructure-as-a-Service (IaaS) providers, such as [Amazon Web Services EC2](https://aws.amazon.com/ec2/) or [Google Compute Engine](https://cloud.google.com/compute/).
An always-on server, such as a university-owned server.
A personal laptop, such as might be used during model development.
The configuration of a model engine for cloud computing using the task scheduler approach is described below.
Configuration of OpenAgua for model engine use
NOTE: For the time being, only administrative users can set up and configure models within OpenAgua for different computing modes, but this will change in the future as OpenAgua is improved. However, non-administrative users can still set up model engines to perform the modeling computations, as explained below.
Under construction.
Configuration of the model engine
Under construction (see WaterLP as an example)
Last updated