Why simplification holds the key to efficiency and productivity
There are few environments more challenging to manage than high-performance computing clusters. HPC centers continually push the envelope attempting to squeeze every last bit of performance from their machines. They use state-of-the-art hardware, advanced interconnects, and fast storage all bound together with the very latest Linux distribution and kernel patches.
Application environments are complex too, with a blend of open-source, in-house developed and commercial software tooling that all needs to function on a common systems software infrastructure. Application tools rely on various versions of system level components like MPI implementations, specialized math libraries, and other software components including workload management systems and workflow automation tools.
Knowledge workers in these environments need to be free to focus on science and not computer science. For example, a biologist working with tools like NWChem, BLAST or BioPig has their hands plenty full just developing and understanding their own models. The last thing they need is to be burdened with learning low-level tool-specific commands or arcane syntax to query job or cluster status.
Some frequent challenges scientists and engineers run into that inhibit productivity are:
Complexity is not a new challenge. Traditionally centers have dealt with this challenge by creating scripts in an effort to automate tasks and hide-complexity from users. Many centers end up with thousands of complex scripts, authored by people with diverse coding styles many of whom have long since left the center. Such scripts are brittle, difficult to maintain and are time-consuming to troubleshoot when something goes wrong. Also, they rarely survive an upgrade to system or application software without needing to be tweaked in some fashion.
Fortunately, there are some good answers to these challenges. In the Platform LSF workload management family, there are two integrated tools designed to tackle exactly these challenges.
IBM Platform Application Center is a flexible, application-centric web portal that can be used to easily create tailored interfaces for cluster users. Such interfaces are essentially self-documenting and are easy to use, deploy, manage and support.
Also, the challenge of complex scripts can be avoided entirely with visual workflow automation capabilities provided by IBM Platform Process Manager. Platform Process Manager enables organizations to automate computational or analytic processes, making them reliable, self documenting, and capturing repeatable best practices without the need for coding or scripting.
By using these tools together to streamline and automate access to tools and data, HPC centers can realize several benefits
You can learn more about IBM Platform Process Manager and Platform Application Center by viewing our webcast at http://ibm.com/platformcomputing/welcome/maximize.html .