This page contains a Flash digital edition of a book.
Give us a job


Workload management has become an essential tool for the HPC administrator, as Warren Clark discovers


As with most areas of HPC, the terminology can be defined in any number of ways, depending on who you talk to, and that’s certainly the case with ‘workload management’. Bill Nitzberg, chief technical officer for Altair’s PBS Works division, would expand the area to ‘HPC workload management and job scheduling’. ‘Tere are three high level goals that


every workload management product wants to achieve,’ he says. ‘Tey are: deliver productivity to the user; allow administrators to set policies about how the IT infrastructure is used; and reduce waste – whether that might be in computers, soſtware, power or people’s time. ‘At the nuts and bolts level, HPC users have work that needs to get done. Tat work


24


is packaged up and sent to the “workload manager”, which receives several packages from different users and basically does traffic control – deciding what jobs should go where and when.’ For newcomers to workload management,


or indeed for those looking to revise their workload management options, there is no easy process by which one can select the best available product. ‘Every site has unique requirements,’ says Nitzberg. ‘In HPC, there is no way that workload management suppliers can come up with a single solution that will work for everyone. Te key is to select a supplier who will truly be a long- term partner and a product that can be extended and customised to meet your specific needs.’


Chad Harrington, vice president of


marketing at Adaptive Computing, adds: ‘Workload managers need to have greater insight into the applications they run. Te more deeply the workload manager can understand the workload, the more efficiently it can schedule and modify the environment. Today’s workload managers understand basic workload requirements and can track an application’s progress. However, there is more that can be done. In the future, we’ll see more emphasis on application insight, including application-specific metrics. If the workload manager understands the applications’ current and future needs for things such as I/O bandwidth, memory allocation, storage space, CPU and GPU cycles, etc., it can make much more optimal decisions.’ Harrington also believes ‘Big Data’


is having an impact. ‘Modern scientific computing operates on massive amounts of data; far more than ever before,’ he says. ‘Managing this data is difficult and the future of workload management depends on its ability to efficiently manage this flood of data. Specifically, Big Data applications need


Arjuna Kodisinghe/Shutterstock


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32