The widespread use of software engineering analysis systems and the need for high-performance resources are driving the development of cloud computing. Performing engineering calculations in the cloud requires solving a number of technical and organizational problems.
Practical tips for creating an effective system for cloud engineering calculations:
1.) Avoid moving large amounts of data
The first practical tip concerns data storage and is to minimize the movement or copying of large amounts of data between cloud resources and the user’s workstation. Obviously, the certain minimum data volume should be transferred: the engineer can create 3D model in CAD system on the workplace, and the given model should be loaded in a cloud.
The size of a file with 3D model for calculation is usually insignificant and takes some megabytes that will demand no more than minute for model loading in a cloud.
On the other hand, the file with the finite element model and the more so the calculation results can take from hundreds of megabytes to several tens of gigabytes. Transfer of such files over the network may take hours or even days, which is not best practice.
2.) Use remote visualization tools
The second tip has to do with the idea of storing data in the cloud. End users need to use high-performance cloud resources seamlessly. This means that you need to configure not only the batch launch of the solvers, but also provide a graphical user interface to work with engineering calculation software – pre and post-processing.
This requires a remote visualization tool with acceleration on the server side and minimal delay. From a cloud resource it is necessary to ensure the delivery of a remote desktop for full work. These factors require a graphic server with enough memory in the cloud to load and display finite element models.
3.) Secure your network connection and storage system
This advice refers to data security – one of the most common issues arising in the design of cloud computing. Calculation models often contain information related to commercial or state secrets. The company needs to be sure that such data is reliably protected.
Simplistically, we can say that the issue of data protection comes down to two factors:
- Installing a VPN node connection, which will require a significant amount of work and will lead to the possibility of scaling – providing multi-user access and work with large amounts of data.
- Perform all data transactions in the web interface through the HTTPS protocol, which is a more pragmatic approach for small groups of engineers or only occasionally using the cloud resource.
Fixed data is protected by encryption. This can be done at the file system or application level. Encryption at the file system level is more easily implemented if good system tools are available.
If you are using an external cloud rather than an enterprise cloud, you must ensure that the cloud provider provides dedicated segments of the system to different clients, isolating your data from other users.
When using cloud systems, participants must share responsibility for security:
- The cloud provider provides physical security for your data center and intranet;
- Cloud user makes sure of the security of his operating system and applications, access rights system, network protocol;
- application developers, in particular, software systems engineering analysis, ensure the security of the software system, in particular encryption during data exchange and storage.
4.) Configure access to the task scheduler and data management system
Many users wonder if the performance of calculating engineers will decrease when using cloud resources. For end users, it is important to have simple, intuitive procedures to run a calculation task that are configured for each specific application.
For the tasks started on calculation it is necessary to carry out monitoring of performance of calculation. To results of calculation access with possibility of archiving, transfer and search should be provided. The corresponding systems can and should be established and adjusted by the supplier of cloud resources.
5.) Use existing software licenses
Licensing issues for cloud computing software products come immediately after data security issues. A key factor for the economic impact of cloud computing is the ability to use existing licenses for engineering analysis software in the cloud. The software provider must provide legal support, and the cloud provider must provide the technical ability to migrate licenses to the cloud.
6.) Consider different licensing models
When using cloud resources, the cost of computing power most often occurs on the principle of payment for the used CPU, GPU cores, gigabytes of disk space. Similar licensing models are offered by some developers of commercial software for engineering analysis. The use of flexible licensing models may or may not be advantageous compared to standard schemes (perpetual or leased licenses).
Where strict budgeting of software costs is needed and it is not possible to determine the amount of calculations in advance, a standard scheme is preferable. In case of peak loads and urgent projects the scheme of payment for licences “on hours” can appear economically more justified.
7.) Balance computational power with tasks
Not all engineering calculations require the same resources and capacity in the cloud. If the solution of a non-stationary problem of computational fluid dynamics with a multi-million grid and excellent scaling requires the maximum speed of calculations, then up to 30 thousand cores can be used for its solution.
Other parameters of a cloud resource may be needed in case of simultaneous solution of hundreds or thousands of mechanics tasks per hundred thousand nodes of a computational grid, in case of parameters space investigation or optimization. It can appear that calculations on graphic cards will accelerate calculation, and can and will not render essentially influence.
Probably the problem will be sensitive to productivity of system of an input-output or to delay in the interconnect. For creation of optimum system it is necessary to lead benchmarking tests – independently or having addressed in the profile professional organisations.
8.) Start small, strive for more.
The most successful cloud computing projects are implemented by organizations that already have experience in high-performance computing at their facilities. In this case, the company’s management can see the economic effect of transferring calculations to the cloud, and the IT department is ready to provide professional support and infrastructure organization for calculating engineers.
Successful companies are implementing cloud computing step by step, using mixed licensing models and regularly evaluating the complexity and impact of these changes.
Cloud computing is contributing to the emergence of new business models for delivering engineering analysis software as a cloud service, which could significantly change the market in the next 5 years and provide engineers and researchers with access to available high-performance computing and analysis resources.
However, not all companies at their business development stages need this technology. In the process of using software systems of engineering analysis, each company should go its own way from calculations on personal workstations to the use of local high-performance systems and then to cloud computing.