Improve computational backend (part of Hardening (o²S²PARC primarily))

[elisabettai]

Continuously maintain the o²S²PARC platform, harden, and increase stability and scalability to accommodate increased usage. For example:

• improved computational backend that ensures reliable and efficient job scheduling. Moving forward, all solver jobs will be scheduled via these facilities, enabling users to choose the hardware on which their jobs should run and providing the ability to inspect and operate (e.g., kill) on the job queue (subject to sufficient permissions). This robust backend will be capable of handling up to hundreds of concurrent jobs, ensuring that even the busiest periods will not cause any disruptions to the service. Furthermore, the backend functionality will also be made available through the API, allowing for integration with external systems that want to launch computational jobs and further expanding the possibilities for users.

Deliverable: Slides as part of the monthly NIH SIM-Core Update meeting (continuous task)

wrike: Y7–(Task) 1b: Hardening (o²S²PARC primarily)

[elisabettai]

Y7Q2 Status update (02/29/2024): Continuously Ongoing Implementation: Presented to the NIH at the monthly update meeting on 02/07/2024

A summary of the key activities performed during the reporting period (presented at the monthly meetings):

• New monitoring infrastructure for the computational backend, via new command line tools to monitor on-demand AWS machines and enabled tracing of created machines by tagging of EC2 instances
• Job concurrency: we strongly improved the optimality of task distribution to machines
• The security of the computational backend has been enhanced by encrypting of data transfer between different components
• The o²S²PARC API server has also been strengthened as a follow up of the milestone DRC and o²S²PARC API Extension to ensure robust real-time streaming of Service logs and metamodeling application
• Monitoring tools were added to the API server to scrape metrics and provide insights into any load issues on the server level. Such issues were identified and resolved.

[elisabettai]

Y7Q3 Status update (05/27/2024): Continuously Ongoing Presented to the NIH at the monthly update meeting on 03/27/2024 (see slides and minutes)

Improvements to the interactive services backend, particularly to reduce start time of interactive services:

• Reworked connection to the cloud filesystem to avoid timeouts while users open their studies
• Improved start time of interactive services on autoscaled machines (i.e. machines created “on demand”), based on an in-depth profiling of all steps and comprehensive optimization. As a result, it is now possible to elastically add new machines and launch interactive services in <1 min
• Extended interactive service backend for accurate progress reporting while pulling a Service’s Docker image #1083