An Introduction to OMF

Control and Management Framework is a pioneering conceptional structure, designed and modeled to enable managerial aspects of testbeds under a unified schema. Deploying different testbeds for experimental purposes might be significantly expensive. Furthermore, users might find big difficulties in describing and setting experiments, uploading and retrieving measurement data in different testbed types due to the lack of a simplified procedure for using and managing testbeds. Taking those parameters into consideration, OMF aims to provide users with ease of use in running experiments, simplified management and optimized control and resource allocation.

OMF Concept

An overview of OMF capabilities. Different experimental platforms under a unified control structure.

 

OMF Concept

OMF Workflow. An illustration of binding together experiments, results, and hardware under OMF.

 

Note. Those pictures are property of [omf.mytestbed.net] and are used under their permission.

 

OMF Components

OMF can be divided in three major components

• Experiment Controller - EC A user sends an experiment description written in Ruby to EC which is responsible for the interpretation and execution of the commands required to the experiment.
• Resource Controller - RC RC is an entity that is deployed on each of the nodes that consist the testbed and is responsible for the control of the node and the execution of the node-specific experiment commands.
• Aggregate Manager - AM AM consists of a set of services that RC can use to manage and execute experiments on the testbed.

How OMF works?

OMF is called to perform three discrete tasks on a testbed.

• Controlling and supporting an experiment.
• Managing the testbed.
• Measuring and retrieving experimental data.

 

Controlling and Supporting an experiment.

Users of an OMF based testbed are able to describe experiments in a high-level language, based on Ruby, and submit it in sequence to OMF. Experiment Controller is the component denoted to the control of an experiment and forwards commands of the respective experiment to the resources (i.e nodes) of the testbed. Then Resource Controller, the component located on each resource, receives and executes commands and sends back to Experiment Controller information and measurements related to experiments. Interaction between Experiment Controller and Resource Controller supports full control over the testbed and a simplified measurement retrieving procedure.

Managing the testbed.

The responsibility for the testbed management belongs to Aggregate Manager. Those responsibilities can be summarized as follows:

• controlling the Chassis Manager card, which enables setting a node on or off (feature soon to be available at NITOS)
• starting and stopping Frisbee, which is used for imaging the nodes.
• enabling nodes to boot either from a network image (PXE) or their local disk
• providing inventory information for all resources
• operating the measurement collection server for retrieving experimental data.

Measuring and retrieving experimental data.

OMF uses the Orbit Measurement Library for collecting experiment results. It gives real-time experiment data collection. Furthermore, this library provides the users with customizable measurement points inside C/C++ application source code. Thus, it gives users the opportunity to specify different measurement points for their experiment analysis.

 

References

• [omf.mytestbed.net]
• [orbit-lab]