RETAIL

Meiosys, a provider of transparent middleware for optimizing and protecting business-critical applications, today announced MetaCluster HPC version 1.5, a software solution for checkpoint and restart (CPR), application migration and resource optimization for high-performance computing environments. MetaCluster HPC provides these functions without the need to modify, re-install or re-link applications. The software provides flexible configuration options for application checkpoint and relocation. MetaCluster HPC v1.2, a single node version, is available now. Meiosys plans to ship MetaCluster HPC v1.5 in spring 2004. In the past few years, high-performance computing centers have adopted commodity-based platforms and clusters to achieve lower computing costs. Nonetheless, many organizations have made a significant investment. These centers are now managing hundreds or thousands of commodity servers, and operators are faced with rising management costs as well as the challenge of providing high availability and high utilization of computing resources. "Technical computing suffers for 'a miss is as good as a mile' syndrome in regards to system reliability," said Dr. Christopher Willard, research vice president, Workstations and High-Performance Systems, IDC Research. "Applications can run for hours, days or months. If an application is stopped a few minutes before completion, all compute costs associated with that application up to that time could be lost. Checkpoint restart is the traditional cure to the syndrome. However, as computer applications and architectures become more complex, adding this capability to each application leads to significant time and programming costs. Thus the ability to provide checkpoint restart as part of the overall system software package is an important requirement for both current and future technical computing environments." MetaCluster HPC is built on a patented infrastructure virtualization layer called MetaProcess(TM). MetaProcess is a transparent technology that dynamically introspects system resources used by the application. It virtualizes Process IDs, Thread IDs, Inter Process Communications, IP addresses and TCP sessions. In essence, the application is decoupled from the underlying hardware and operating system, offering IT managers a way to manage resources effectively without having to learn new applications or disrupt ongoing operations. "We believe we are the first company to provide transparent checkpoint and restart functionality for standard operating systems, legacy applications, commercial (COTS) and custom applications," said Marc Rougier, President and CEO of Meiosys. "IT managers using MetaCluster HPC are, in effect, getting an insurance policy for long-running jobs. In addition, they can optimize their clusters by dynamically relocating applications to faster systems when available, or implement more flexible license release policies to efficiently support multiple projects." Features and Benefits MetaCluster HPC offers several features that provide cost and time savings benefits: -- Transparency -- MetaCluster HPC delivers checkpoint and restart, application relocation and resource optimization without service interruption. -- Holistic checkpoint mechanism -- MetaCluster HPC captures the full context of an application such as memory usage, file status and any open TCP/IP connections, leaving nothing to chance. Checkpoint files can be stored locally or on remote or shared storage within the network. -- Application protection with periodic checkpoints and automated restart -- IT managers can "set it and forget it," virtually guaranteeing long batch runs. -- Integration with existing batch managers -- Meiosys seamlessly operates with leading batch managers such as PBS Pro and Platform LSF, eliminating the need to learn new user interfaces. -- License usage optimization -- With MetaCluster HPC, applications can be checkpointed and licenses immediately released. For example, engineers from one department may borrow licenses owned by another department when not in use, and upon request, checkpoint their own jobs and release the licenses. Checkpointed jobs can be resumed at a later time when the licenses become available. Large scope of supported applications MetaCluster HPC supports a range of clustered and traditional HPC application types, including single and multi-process, single and multi-threaded, sequential and distributed (e.g., OpenMP, MPI), legacy, off the shelf and custom applications. MetaCluster HPC supports 32- and 64-bit Linux installations (RedHat 7.2, and above). The software is integrated with Platform Computing's LSF batch manager (4.0.1 and above) as well as Altair Engineering's PBS Pro (5.4). Meiosys plans to support Solaris in future versions of MetaCluster.