Long-Term Campaign
According to the IRI Architecture Blueprint Activity [B4], long-term Campaign patterns are characterized by a need for sustained access to resources at scale over a longer time to accomplish a well-defined objective. Robustness, reproducibility, and reliability are important to accomplishing long-term science, and these patterns will likely involve significant logistical planning. Examples include sustained simulation production and large data (re)processing for collaborative use.
The Long-Term Campaign Pattern Group of the IRI Architecture Blueprint Activity determined that the key overall challenge is to intentionally plan and coordinate resources between campaigns and facilities over time. Specific challenge areas include long-term storage past the end of a project; a present mismatch between the short-term nature of resource allocations and mechanisms (i.e., compute and instrument time) versus the long-term needs of a campaign; the continual evolution of technologies and approaches within facilities and campaigns (e.g., computing architectures, infrastructure and instruments, cybersecurity, workflow systems); avoiding interruptions in campaigns due to facility downtimes; and the present lack of holistic approaches to resource allocations across all DOE-SC user facilities.
Additional perspectives from the group included:
The evolution of staffing duration of a campaign needs to be factored in.
The facilities may provide infrastructure and also need to accommodate the varied data management requirements of the programs and research domains.
Abstraction layers are likely to be a key and pervasive component of the solutions we need.
A common machine-usable interface to facilities looks like it is a prerequisite, as is scheduling that does not have humans in the loop.
Data re-use is an appealing idea but difficult to achieve, so data needs to be well-described and documented to be useful down the road.
Relationship to INTERSECT Science Use Case Design Patterns
The Long-Term Campaign IRI Pattern is related to all INTERSECT patterns in terms of (1) long-running experiments or (1) sets of long-running experiments.
This IRI Pattern is more specifically related to the INTERSECT Experiment Control strategic pattern and its Local Experiment Control and Distributed Experiment Control architectural patterns, when a single experiment needs to be performed over a long period of time. For example, an open-field plant growth experiment investigating the absobtion of CO2 may run for months and even years, collecting experiment data in regular intervals. The resources required to control this single experiment need to be available over an extended period of time.
It is also related in a similar way to the INTERSECT Experiment Steering strategic pattern and its Local Experiment Steering and Distributed Experiment Steering architectural patterns, when such a single, long-running experiment needs to be steered, i.e., its parameters need to be changed over time based on experiment progress. For example, an open-field plant growth experiment investigating the absobtion of CO2 may regulate the amount of CO2 in different parts of the field to adjust for environmental factors, such as sun exposure and rain. The resources required to steer this single experiment need to be available over an extended period of time.
The Design of Experiments strategic pattern and its Local Design of Experiments and Distributed Design of Experiments architectural patterns are related to this IRI Pattern, when a set of similar experiments with different paramaters is performed over a long period of time. A parameter study in search of a new material for microelectronics that performs in a specific way, such as to create a light-emmiting diode in a certain wavelength, may go on for a prolonged period of time. Each new experiment is a modified variant of the previous one, potentially using different materials or techniques for depositing layers of materials on semiconductors. The resources required to design and control such a set of experiments are required over an extended period of time.
This pattern is also related in a particular way to the INTERSECT Multi-Experiment Workflow strategic pattern and its Local Multi-Experiment Workflow and Distributed Multi-Experiment Workflow architectural patterns, as multiple experiments that may depend on each other are performed over a long period of time. There are two types. In the first type, a workflow is repeatedly executed over a long period of time, such as when plant samples are regularly taken from an open-field plant growth experiment investigating the absobtion of CO2 and analyzed in a chemistry laboratory. The second type involves long individual experiments that are part of a workflow, such as the creation of an isotope in a reactor over a month, the investigation of its initial properties, and the investigation of its properties after decay every month. The resources needed for both types need to be available over an extended period of time, where individual resource availability differs between types.