We’re happy to announce Moab Task Manager, a localized decision-making tool within Moab’s HPC Suite that enables high-speed throughput on short computing jobs.
“Moab Task Manager is very valuable to customers who want drastically improved throughput on production clusters that have many small, short-lived, similar jobs,” said David Jackson, Adaptive Computing’s CTO and co-founder. “When pushing a high number of jobs through the system, Moab Task Manager works in unison with Moab to facilitate parallel job scheduling with larger, longer-running jobs.”
Capable of running in Moab and TORQUE environments, Moab Task Manager is capable of launching 10 jobs per node per second for up to 100 times faster throughput, all while reducing latency. This results in greater efficiency when executing short jobs, which are very common for today’s workloads, running for a handful of seconds and requesting few resources.
By taking the bottleneck of launch speed out of the equation and preventing logjams and delays, Moab Task Manager is the software ticket to exascale computing. Moab Task Manager’s speed results from its capability to incur scheduling overhead only once for a large batch of jobs, instead of once per individual job. Thus, a Moab Task Manager job batch obeys traditional policy and scheduling constraints when preserving Moab’s power and flexibility.
Moab Task Manager can be used on existing clusters with no Moab or TORQUE upgrades or environment modifications. Following its first release, Moab Task Manager will continue to undergo architecture enhancements, including deployments with other scheduling technologies.
Want to see Moab Task Manager in action? We’ll be demonstrating it at Supercomputing 2013 in booth #3113. Also, representatives from Adaptive Computing’s management team will be participating on the panel, “Programming and Managing Systems at Scale” to discuss how refining architectural techniques will provide the necessary foundation to achieve exascale computing.