Application Performance Management (APM) for HPC

Everybody needs APM (application performance management): that’s the high-level message that its zealots, like Aberdeen Senior Research Analyst Jim Rapoza, rightly repeat. That universality can obscure, though, how APM fits in particular situations–especially in an information technology (IT) world where we all think our requirements are like no one else’s.

Consider for a moment the HPC (high-performance computing) domain. Mary E. Shacklett, President of Transworld Data, points out that APM originated with a focus on OLTP (online transaction processing), where the pertinent metrics are much different from HPC’s concern with parallelized computational efficiency.

The business goals are plenty familiar, though. As Shacklett puts it, “… management software can provide workload monitoring, reporting, and management … so system administrators can tune for optimal performance for users while also optimizing resource utilization …” This is particularly timely as HPC advances from the craftwork of hand-coded, often rather amateurish applications and operations toward “best practices” that can be replicated and managed more professionally.

That transition is evident in the HPC world’s interest in “app stores”, cloud architectures, other virtualization techniques, and business-oriented trends. nanoHUB, the NSF’s (National Science Foundation) HPC community showcase, exemplifies the possibilities in its emphasis on re-use, collaboration, and standardization. At an operational level, advanced HPC practices give “… utterly dependable operation, low-cost and rapid content adaptation and deployment …“, among other benefits.

HPC is a specialized and even esoteric corner of IT, where conventional tools often fit poorly. Even there, APM’s landmark benefit–swift and reliable diagnosis of performance anomalies–makes for good returns-on-investment (ROI). Next week, “Application Monitor” will return to show why APM is also a crucial priority for the IT mainstream.