ANSYS, Lenovo and OCF have joined forces to put a turnkey HPC appliance into the hands of the UK engineering community. The plug-and-play appliance, designed to run computational fluid dynamics (CFD) simulations, is the result of a unique partnership between engineering simulation software specialist ANSYS, hardware vendor Lenovo and high performance cluster integrator OCF.
The ANSYS CFD-ready appliances are pre-built with a fixed head node, chassis, memory, the latest Intel processors and fixed switches. They are configured and tested with ANSYS software and scheduling software to ensure smooth deployment. The use of blade architecture means that they can be easily scaled to fit customer needs.
Even the smallest option in the portfolio, a cluster with two dual-socket Lenovo NeXtScale compute nodes, offers twice the compute performance of most workstations and with intelligent scheduling enabling jobs to be run around-the-clock, engineer productivity can easily be six times that of a single workstation, according to OCF. Appliances are fully supported by the system integrator, so customers are covered if there is an issue with the scheduler issue, network or minor disk failure.
This ties into the reason why a firm would choose an appliance versus building it themselves.
“The benefits of the solution are that a lot of the architectural decisions are already made for the customer,” OCF’s Andy Dean shared with HPCwire in an email. “Things like: what network do I need? What software do I use to manage the system? Which scheduler should I use? Which processors should I go for? Do I need local scratch disks? OCF has worked with Lenovo and ANSYS to make these decisions to build a system that offers the best compromise of Usability/Performance and Cost.”
“We then build this environment in-house delivering a pre-racked, pre-cabled system with the base software installed ready for final integration into the customers’ environment and workflow – drastically reducing the time from purchase to first simulation, when compared with a customer buying ‘a load of boxes’ and building their own system.”
The scalability that a cluster provides to engineering codes means that customers can run larger, more complex simulations or get more simulations completed. Ultimately this leads to better product designs and/or faster time to market. Dean reported that many of ANSYS’ customers have tried to run jobs over a couple of workstations and not had the best experience.
“Clustering is complex and most ANSYS users are engineers not IT or HPC engineers,” he added, “so this partnership brings HPC experts (OCF/LENOVO) to ensure these HPC projects are success leading to better simulation and better products.”
When asked about their vision of the ideal customer, Dean explained that industries that use simulation for product development run the gamut from ‘crisps to cars.’ “So it can be any industry,” he stated. “But as a general rule if a customer is using between 4-6 workstations (or more) in their organisation for simulation they should at least consider the potential benefits of using a HPC system like the Lenovo solution for ANSYS. This is only a general rule though – I’ve seen one-man outfits with huge, complex, simulation requirements.”
As for pricing, it depends on system size and the software selection. “Some customers use ANSYS Mechanical, others CFD codes, and the variation in size and complexity of the simulations potential customers have is huge,” Dean shared. “That said, a small cluster would start at £30-40k [$44-59k] which is a significant investment, but by comparison, you would by surprised by the number of £5-£10k [$7-15k] servers engineering firms are purchasing to try and keep on top of their simulation requirements, and the potential improvements in efficiency that can be realised offers a compelling business case for heavy users of simulation.”