Over the last couple of years, more and more companies are wrestling with the same question: “Shall I continue to invest in on-premise hardware or switch to cloud computing?” I usually respond: “Well, it depends. …” Because it depends on their business, use cases, infrastructure and other factors, I advise them to carefully consider eight important questions.
How important is business agility for your organization?
Being able to expand compute resources almost immediately is no doubt a clear advantage of the cloud. With a solution from an ANSYS cloud-hosting partner, you can have a complete, turnkey simulation data center capable of serving your simulation needs in one-to-two days, instead of the months it takes to plan, order, install, provision, configure and test an on-premise solution.
Working in the cloud, you can also easily add new engineers. You don’t need to expand your HPC capacity or buy these users powerful new hardware: You just provide them with a login to the cloud portal. You can also start with a small, specific project and branch out as you gain confidence with the cloud, or as your on-premise resources are gradually decommissioned. Finally, you can also grow to serve global business needs by extending access to the same, consistent simulation platform to engineers in other geographical regions.
Do you have significant variations in your simulation workloads?
The ability to scale up and down your computer and license capacity is another clear advantage of the cloud. The benefit here is that you get the right (i.e., ANSYS application-specific) HPC resources you need, as much as you need, when you need it, and pay for only what you use in compute and license capacity. When an HPC job is submitted, the job runs without the long queue wait times that may happen with on-premise solutions. And, when the job is finished, the cloud computing servers can be shut down.
When you know your exact simulation workloads and don’t foresee any big variations in workloads, having access to capacity will likely be less important for you. In this case, the most cost-effective choice would likely be a dedicated, on-premise server and traditional lease or perpetual licenses.
Does your organization have Capex constraints?
With on-premise computing, there are usually higher upfront capital expenditures (Capex and operations and upgrade costs, compared with cloud deployment requiring less infrastructure and support staff. With the on-premise option, you will need long-term planning and commitment of resources for scaling. In contrast, cloud solutions can be easily scaled up and down with little time and effort.
Cloud computing is an effective way to replace Capex with operational expenditures (Opex), and ideal for organizations with significant swings in engineering simulation activity. Moving to the cloud may nevertheless require a new information technology (IT) budget discipline and skill sets. Usually with the cloud, you get higher year-over-year maintenance costs compared with on-premise deployment. But these higher costs could be well justified if the cloud solution delivers more major benefits, including speed in product innovation, a marked increase in business agility and a continuously honed competitive edge.
How important is it to your organization to maintain a sense of control? I can mention here that corporate culture is likely an important part of the decision for maintaining the status quo or moving to off-premise computing. If a strong bias exists for keeping resources and expertise in-house, the easiest (but not necessarily the best) choice will likely be an on-premise solution.
Do you have plans to enlarge the geographical footprint of your organization? If this is the case, you’ll be more agile with a cloud solution and able to expand compute resources almost immediately. And, with cloud-hosting partners supporting global datacenters, you’ll have access anywhere in the world.
What’s the level of IT support available to you? If the level of IT support is minimal or nonexistent in your organization or you don’t want to involve your IT department in deploying HPC resources, you have two appealing options:
If you can involve your IT department, then they will likely assume responsibility for hardware maintenance, availability of resources, software updates and disaster recovery.
Does your organization need to address big data analytics and/or internet of things (IoT)? The adoption of big data analytics and IoT compels the move toward cloud technology. According to International Data Corporation (IDC), “Within the next five years, more than 90 percent of all IoT data will be hosted on service provider platforms as cloud computing reduces the complexity of supporting IoT ‘Data Blending.’” If your organizational needs include big data and IoT, you will naturally want to invest in the cloud for engineering simulation.
What’s the level of customization required?
The cloud tends to be more rigid when it comes to customization —many cloud partners offer options (but for additional fees). And, there is usually greater stability and ongoing upgrades as a result. With on-premise, you will likely get the most flexible option for customization because your organization controls the solution. There is, however, a downside to this too: It may lead to problems in terms of implementation time and issues with software updates.
Let me conclude my blog with a few takeaways. First, cloud computing is becoming imperative as more and more customers are looking for ways to scale up HPC to run a greater number of faster, more complex simulations. Second, you need to ask yourself the above questions, weigh the options in light of your organizational goals, consider the differences between the cloud versus an on-premise solution and determine which is better for your organization and why. And last, but not least, on-premise and cloud choices aren’t mutually exclusive; it is obviously possible to have both!
I would be very interested to hear about your thoughts on cloud versus on-premise computing, and on other key questions you’ve been asking yourself. Thanks!