The Workload Knows

A workload is a dynamic virtualization that encapsulates the tasks a system must perform under varying and often unpredictable conditions to meet its design requirements.

Workloads are one or more of the functionalities that exist internally to a system or more broadly described externally as the application space that consumes internal system resources and influences metrics such as energy use, response times and performance outcomes.

Developers know that classical configuration practices are all about system resource optimization for meeting the demands of their workloads. Workloads “know” their status at every instance and the developer’s job is to anticipate how best to resource a system to optimally meet those requirements regardless of how the workload operationally presents itself to the system as built.

During the development process, at any managed interface, workloads are simulated (and often inferred) from experience and data, but always have at any time the potential for unpredictability, just as hardware and software system resources can sometimes be unpredictable as well. 

This relationship between workloads and the system resources managing those workloads is the source of all configuration complexity and conflicts, yet even understanding this does not resolve the dilemma that a system and its workloads can never be perfectly in harmony, all the time.

The System is the Workload

In Self-Aware™ and AdaptiveAI™ computing, workloads are the driver to dynamically setting system configuration parameters that are responsive to any conditions of the instant. Workloads are descriptive of the sum of all conditions that exist at that instant and configuration settings are continuously, dynamically managed to meet prescribed goals under software control.

With this dynamic computational capability, workloads too difficult to “anticipate” or manage when only using conventional practices, can now be targeted for optimization with their own, unique and meaningful goals to modify any parameters in the exposed configuration space made responsive to control that interface. 

The basic paradigm to enjoy these enhanced capabilities, available from Config, include: 

1. Conceptualize any hardware or software system as a multiplicity of workloads and identify every interface of complexity within a system, typically at the nexus of various o/s and software packages being integrated. Develop a preliminary list of primary goals for each interface. 
2. Optimize your system’s configuration space parameter settings by analyzing the resources your system requires over its operating cycle based upon demands of the workload. 
   • (Basically, configuration parameters are made variable, a variety of goals are established for evaluation and the workload is run to create dynamic models of potential resource applicability and requirements.)
3. Use the tools of our Sequitur™ Platform, or with guided assistance from our system integration team and establish API connections to our computational methods. Add goals to your existing system, launch, collect performance data on actual workload operations and reset goals accordingly to map to achieving your performance objectives. 
4. Build the system you normally would build, following your own design practices and tools, to your own design specs and standards.
   • (Developers may also consider using the original design specifications as initial, most hierarchical goal for dynamic management of the complex interface between the as-built system in its entirety and the instant-by-instant experience of the real-time running of the workload.)