When running engineering and scientific simulations, high performance computing (HPC) resources are often vital components in successful analyses being executed in a timely manner. The multi-billion dollar HPC industry continues to grow to meet popular demand for the improved performance, faster interconnectivity, and lower network latency that HPC computing resources offer over standard computing configurations.
The growth of the HPC market includes, not only traditional on-premise clusters, but also cloud HPC. Cloud HPC has experienced, and is estimated to continue to, achieve steady growth alongside the booming HPC market. Research estimates that by 2015, cloud spending for the technical sector will be $235.1 billion–cloud architecture in 2011 was $78.2 billion. For the cloud HPC market specifically, analysts note that it is growing at a rate similar to the general HPC market. The general HPC sector is estimated to grow to $30.2 billion by 2018—up from $21.9 billion in 2012. Of that $21.9 billion spent on HPC, 23.5% was spent on cloud HPC applications.
With the increased demand for cloud HPC, there have emerged many providers who offer various HPC solutions in the cloud to engineering and science companies. But how do potential users of cloud HPC navigate the uncertainty and questions that surround running proprietary analyses in the cloud? Questions and concerns vary from industry-to-industry, company-to-company, and user-to-user; however, below are a couple of important questions to ask prospective cloud providers when determining whether they are a good fit for your workload.
What software tools are available?
Supported software is a critical differentiator. Some cloud HPC providers only support open source tools, while others work with commercial software vendors to support licensed software. While some software vendors offer licensing in the cloud, other environments support the ability to use existing software licenses at little to no extra software cost. As a user of simulation software, it is also important to understand what software functions are available in the cloud.
Security is an important topic when considering cloud simulation and can include multiple departments within a company. A few specific questions to ask include:
- What is the level of data encryption?
- Does the cloud provider submit to external security audits?
- Do users’ simulations run on isolated clusters?
- What are the redundancy procedures?
- Is there an International Trade in Arms Regulations (ITAR) compliant environment? (for military and defense companies)
There are certainly many other questions to ask, and IT teams often have their own company specific questions for cloud providers.
What are the costs and fees for running in the cloud?
Many providers offer a pay-as-you-play usage model while others offer monthly subscriptions or annual payments. Different models will be cost effective for different workloads. Ask cloud providers what is included in the price and whether any additional charges may apply. Often times there are costs for hardware and separate costs for software, storage, data transfer, and visualization.
Can I test your service before I commit?
With all of the different options for cloud HPC available, it’s important to find the best fit for your workload. Several cloud providers offer the opportunity to test out their environment at a free or reduced price before committing. This will allow you to test not only a cloud HPC provider’s resources, but also their service and support.
What are the data management and collaboration features?
The primary function of cloud HPC providers is often to reduce the time it takes to run computationally-intensive simulations. However, beyond that, there are many other features that can significantly enhance your experience when running simulations in the cloud. These include the ability to easily share information with colleagues, the ability to copy simulations to make quick iterations, the speed of data transfers, what visualization capabilities are available, and what customer support is offered, amongst others.
Using cloud HPC resources for CAE simulations is becoming more widely adopted, and according to research, estimated to continue to grow rapidly. Testing the simulation environment and asking questions is the best way to determine which platform will work for you.