Altair has announced the general availability of PBS Professional 13.0, the latest version of its software for high-performance computing (HPC) workload management and job scheduling on clusters and supercomputers.
The new version has been designed to meet the needs of companies looking to prepare for the move to exascale and represents the largest release in PBS Professional’s 20-year history as a commercial product. Its new features and capabilities include faster job launch, a comprehensive health check framework and scalability to millions of jobs per day.
‘For this release we put a great deal of focus on speed, scale, and resilience,’ said Bill Nitzberg, CTO for PBS Works at Altair. ‘We re-architected the communication layer – making it multi-threaded, non-blocking, persistent, and fault-resilient, with asynchronous job starts – delivering ~ 10x faster throughput, ~ 5x greater scalability, and a 100 per cent health check framework. Not only does this make PBS Pro ready for exascale, it greatly benefits all our customers.’
PBS Professional is designed to improve productivity, optimise resource utilisation and efficiency, and simplify the process of workload management for HPC clusters, clouds and supercomputers. PBS Professional’s powerful job scheduler automates workload scheduling, managing, monitoring, and reporting, and is the trusted solution for complex Top500 systems as well as smaller cluster owners.
New features in version 13.0 include:
Million-core scalability – tested to more than 50,000 nodes
Fast, reliable startup of huge MPI jobs – tested at tens of thousands of MPI ranks, which minimises delays caused by faulty nodes
Fast throughput on short jobs – supports more than 1 million jobs per day
cgroups eliminate resource contention – jobs run faster and don’t interfere with each other or the OS (limited availability)
Comprehensive health check framework monitors users’ health check script behavior – either checks run or node is marked down
Expanded hook events for more plugin extensibility and customisation
More policy controls to better match clients’ business needs
Fine-grained targeting for preemption, configurable at the queue level (admin controlled)
General fairshare formula enables accruals per-Q, license sharing, time-of-day, power use, and even combinations of these
Expanded support: Intel MPI and MPICH2 on Windows; UNC paths for stdin, stdout, and file staging on Windows; SLES 12 and RHEL 7
Custom resources can be created directly using qmgr, without the need to restart the server
Long job and reservation names supported
Carlo Cavazzoni, HPC Operational Activities and Production, CINECA said: ‘PBS Professional 13.0 is an important release for CINECA – we are excited about the improved job throughput and improved communications, as well as the new capabilities for cgroups. We rely on PBS Professional for secure, scalable management of our HPC infrastructure, and we look forward to installing the 13.0 GA version.’
Simon Burbidge, HPC Manager, Imperial College London said: ‘As we’ve increased the size of our system, the number of jobs we run per day has increased dramatically. We have people running lots and lots of very small jobs doing parametric studies and things like that, so we’re looking forward to the increased capacity of PBS Pro [13.0] to run even more jobs per day.’
Current Altair customers may begin using PBS Professional 13.0 immediately under the same patented token-based licensing framework employed by PBS Works, HyperWorks and other Altair applications. It will also be available as part of HyperWorks Unlimited, a state-of-the art, managed private cloud solution; within HyperWorks On Demand, Altair’s public cloud for computer-aided engineering (CAE); and as part of Altair’s Simulation Cloud Suite for simulation lifecycle management.
Altair will demonstrate PBS Professional 13.0 at booth #1110 during the ISC High Performance conference in Frankfurt, Germany later this month.
