Share this post:
Among organizational leaders operating high-performance computing environments today, there is rapidly growing interest in being able to leverage cloud computing resources on demand to augment on-premises resources – to address peaks in demand or to ensure time-critical work can be completed on time. With more than two decades of experience managing clusters and clouds, we are pleased to announce the release of the third update to IBM Spectrum LSF 10, featuring updates in three main areas which make it easier to gain the benefit of a hybrid cloud approach:
- We’ve made hybrid cloud even easier by further enhancing the IBM Spectrum LSF resource connector capability, which provides autoscaling in the cloud. New bursting policies, along with support for spot pricing and federated identities, provides administrators with even greater control in how cloud resources are provisioned to meet business service-level agreements (SLAs). In addition to Amazon EC2, support for IBM Bluemix and Microsoft Azure have also been added.
- Support for graphics processing units (GPUs) in IBM Spectrum LSF was first added in 2009. Back then the management capabilities of GPUs were simplistic, as were the use cases. Over the years, the usage of GPUs to accelerate high-performance computing workloads has grown dramatically, and we’ve continued to evolve our support for them. In the last year, there has been a huge focus on GPUs for AI. In this update to IBM Spectrum LSF, we have further evolved our GPU support to provide users with even greater control of how GPUs are used with their applications.
- The previous update to IBM Spectrum LSF 10 focused on simplifying the use of container environments with IBM Spectrum LSF. In this update, we have added support for NVIDIA Docker. Administrators are now allowed to dynamically insert site-specific options into the container start up sequence. For example, based on the user and job attributes, specific file systems can be mounted within the container.
Together, these three capabilities make IBM Spectrum LSF 10 an ideal platform for supporting a diverse range of workloads – from traditional HPC and high-throughput workloads, to big data (Hadoop, Spark), machine learning and deep learning (TensorFlow, Caffe) whether on bare metal, virtual environments or containers.
Of course, there is a lot more to IBM Spectrum LSF V10.1.0.3 than just these three capabilities. Find out more about IBM Spectrum LSF here.
New to workload management? Learn about how the powerful cluster virtualization capabilities of IBM Spectrum Computing products can help to optimize the performance and cost efficiency of your infrastructure while simplifying management.