Hewlett Packard Enterprise (HPE), today for the first time in five years, made two additions to its portfolio of fault-tolerant Nonstop computing platforms for on-premises IT environments.
The entry-class HPE Nonstop Compute NS5 X5 and the flagship HPE Nonstop Compute NS9 X5 are based on Intel Xeon Bronze 3400 and Intel Xeon Gold 6400 series processors to double the memory capacity (8TB) of previous generations and 2.5 times more network bandwidth.
Overall, the systems, according to internal HPE benchmarks, provide 15% more performance capacity, with the HPE Nonstop Compute NS9 X5 being able to scale up to as many as 4,000 processors.
Casey Taylor, general manager for HPE Nonstop, said that at a time when more data-intensive workloads involving, for example, artificial intelligence inference models, are being deployed in on-premises IT environments, there is a clear need for increased capacity.
The HPE Nonstop Computing platform traces its lineage back to Tandem Computer, which was acquired by Compaq Computer in 1997. Subsequently, Hewlett-Packard acquired Compaq in 2002, which became part of HPE when it was spun off in 2015. Those platforms enable IT teams to run duplicate copies of application workloads in parallel on two different systems to ensure maximum levels of uptime.
Since then, HPE has redesigned these fault-tolerant systems in a way that provides that capability using a disaggregated software fabric rather than relying on a physical interconnect between two systems. That approach enables HPE to offer an edition of its Nonstop platform that can run in the cloud or an on-premises IT environment. There are also editions of the platforms that can be managed by HPE using the HPE GreenLake service.
It’s not clear to what degree IT organizations are investing in fault-tolerant computing platforms, but with more workloads starting to be deployed in on-premises IT environments, there is an increased need for high availability, especially in, for example, payment processing applications or manufacturing environments, noted Taylor. As a result, HPE is recommitting to its Nonstop platform at a time when the cost of downtime is simply too high not to invest in a fault-tolerant platform, she added.
Fortunately, the cost differential between a fault-tolerant system versus investing in traditional x86 servers is not as high as it once was. There is still a premium, but for some classes of applications, the cost of downtime far exceeds any additional infrastructure costs. The challenge, as always, is explaining that value proposition to business leaders who don’t always appreciate how often server components can fail as utilization rates increase. In some cases, simply reducing the stress level of the IT teams and application developers tasked with maintaining mission-critical applications justifies the expense.
Hopefully, there will come a day when IT infrastructure based on Intel platforms is inherently fault-tolerant to the point where there is never any unplanned downtime. In the meantime, however, the next time a mission-critical application does become unexpectedly unavailable, the cost of the fault-tolerant system will most likely appear trivial in comparison to the actual business impact.
