Hyperion Research conducted a survey for Panasas to gain better insights into the relationship between total cost of ownership (TCO) of High-Performance Computing (HPC) storage systems, the initial acquisition costs and the benefits users derive from greater simplicity in their installations. The survey targeted HPC data center planners and managers, HPC storage system managers, purchasing decision-makers, key influencers, and HPC storage system users. Hyperion, which published their findings in a report, “The Importance of TCO for HPC Storage Buyers” also presented their findings at the Panasas User Group1 meeting.
The Hyperion survey report concluded that when it comes to the purchase of HPC storage systems, certain data, such as I/O performance and the cost of acquisition, are easy to measure. Yet some of the harder things to measure are often overlooked. Specifically, the ongoing costs of operations and the negative impact that inconsistent and complex storage solutions can have on productivity and time to quality outcomes.
Survey participants were asked to identify their most important purchasing criteria and their responses are shown in the graph below. It is no surprise that performance topped the list for HPC storage by far (performance is the field’s middle name after all). And as expected, purchase price and TCO tied as the second-most important purchasing criteria. No news here. Indeed, HPC storage purchases have been driven by performance and price/performance for decades.
Most Important HPC Storage Purchase Criteria
It’s also interesting to note how far down the list “Number and skill of staff required”, “Recovery/resilience”, and “Software support” show up. I have to admit that I found the low ranking of these criteria disappointing as an employee of Panasas. Ours is a company that is uniquely focused on low-management requirements, super-resilience and great support – along with leadership performance and price performance. Since the HPC organizations surveyed appeared not to care about these criteria, some might be tempted to interpret the survey as a repudiation of the considerable time and effort that we’ve invested over the years.
But when we dig deeper, a different, more nuanced picture emerges. When asked what their greatest operational challenges were, HPC organizations puzzlingly responded with issues similar to the list of the lowest priority purchase criteria, as shown below.
Most Challenging Storage Infrastructure Operational Aspects
From their perceived low priority as purchase criteria, it can be concluded that these operational issues appear to be under-appreciated by most HPC organizations (or else why wouldn’t they prioritize them higher?).
Oh, the Irony!
Isn’t it ironic that the buying criteria ranked among the lowest in the survey would pop back up and rank among the most worrisome operationally for HPC organizations? It’s as if buyers of HPC storage only prioritized performance and price without any consideration of operational issues. By extension, subsequent purchases seem to have no memory of operational headaches resulting from previous purchase decisions.
From these results, one might assume that most HPC organizations seem to have accepted that they need to trade off reliability, easy management and good support to get high performance at a good price. Or perhaps they see these criteria as having little impact on their HPC operations. Or both?
What HPC Organizations Fail to Understand About their Operations, But Should
The next few blogs will test these assumptions and their impact, with data extracted from the Hyperion survey. So stay tuned.