Keeping the Vortex spinning: How Galaxy Australia ensures jobs run smoothly
When a researcher uses the Galaxy Australia platform for their data analysis, they expect a straightforward job launch, easy tool selection and a quick response time. This seemingly simple task requires a complex set of resources and a sophisticated process behind the scenes. Total Perspective Vortex (TPV) was designed locally to streamline the management and routing of jobs and led to such massive efficiencies that it has been rolled out around the world.
TPV is able to dynamically set resource requests and route jobs to appropriate compute resources based on a set of configurable rules. By significantly lowering the manual intervention required for administering the system, managing and routing complex jobs in high-throughput environments become significantly more efficient. When it was deployed to dispatch jobs across Galaxy Australia’s highly distributed compute resources, it nearly doubled the maximum throughput.
One challenge administrators face is the huge diversity in jobs that Galaxy Australia users run. With TPV, a user’s choice of tool and their type of data is passed through a decision tree to ensure that the job is appropriately resourced, while balancing the competing demands of job turnaround time, throughput and overall efficiency.
Another strength and complexity of the Galaxy Australia system is its distributed network of computers. Galaxy Australia’s compute infrastructure is spread across a number of national cloud services. Underpinning the simple graphical user interface that users interact with are resources provided by AARNet, University of Melbourne, QCIF, Pawsey, NCI and the Microsoft Azure Cloud. The following table shows the sort of jobs the Galaxy Australia TPV distributes across these resources.
Examples of how TPV distributes jobs to compute infrastructure
TPV is authored by Software Engineer and Research Fellow Dr Nuwan Goonasekera, Melbourne Bioinformatics, University of Melbourne, with support from the global Galaxy community. It is evidence of the constant systemic improvements that are possible when an international community has a shared ambition and another example of Nuwan’s contribution to Galaxy’s international code base.
TPV has a domain-agnostic design that can be adapted to other complex resource management systems. If you are interested in technical details, ARDC is hosting an online information session on 26 Oct: TechTalk 23: Galaxy – Distributed Computing Through Total Perspective Vortex.