Cluster Design Services 2 of 3
- Of course, the most perfect technical solution might not fit your budget, so we'll need to work with you to understand what your desired budget is and any possible trade-offs you may have to make to meet that budget. Your sales engineer will ask you for your desired price range, as well as helping you explore the most critical or highest priority requirements you think you have. But we don't stop there. In some cases, you may not have thought of particular design considerations that might adversely affect your long term reliability or usability. We'll point these types of issues out as we encounter them so that you can take them into consideration.
- Once we have identified your basic needs and wishes, we will generate a quote based on that design. That quote is to be used as a template for your design, and your sales engineer will work through this quote with you, helping you to decide what is critical and what your trade-offs are. Can you do with a bit less storage, or is your performance gain on one particular architecture worth the cost per node difference? Does the use of a particular commercial compiler add enough performance to the code to justify its cost? Do you truly need a non-blocking Interconnect fabric, or can money be saved there? Can you trade space usage (form factor) for lower cost? These things and more are all considered by your sales engineer as part of the design process. You may receive multiple quotes as we fine tune the configuration and work toward your desired solution for deployment.
- During the design process, we will be discussing different technologies that may help your particular situation. We provide a Detailed Buyers Reference for those who wish to explore some of the common HPC technologies farther, and may refer you to specific sections of that document in order to help explain aspects of the technology we're discussing at the time.
- At Aspen, we believe that we're building a solution together. This means that any additional information we can impart can only help you define the most reliable and cost effective long term solution, and that we need your input to truly understand your requirements and situation. We want you to purchase your next cluster from us, and the next one after that, so we're not just worried about what you think of the cluster you buy today, but also what you think of that same cluster after you've used it for three, four, or even five or more years. To ensure that your future experience is as trouble free and enjoyable as possible, we go the extra distance and look at and point out your long term requirements, something that many HPC vendors do not do. Your sales engineer will be thinking of ease of maintenance, total cost of ownership, the best price performance ratio for your needs, and many other issues as he or she works with you on budget and requirements. Your sales engineer will also be asking you questions and thinking hard about your particular environment.
- Clusters don't function in a vacuum, they have to work with your local environment. Your sales engineer will work with you to determine what power voltage you have or will retrofit into your facility, what PDUs will work best in your environment, what plug types you require, and whether you require UPS power protection for one or more of your nodes.
Depending on the size of your cluster, power and cooling requirements can be much higher than customers may think, and these capabilities must be sized appropriately for your cluster to live a long and happy life at your site. In some cases, the required infrastructure doesn't exist, or will be too expensive to retrofit, which is why Aspen offers Add-On Service options such as HPC Co-location. In other cases, your cluster will need to live in a facility that was not originally designed to house a modern cluster. This is why Aspen offers cooling solutions to solve your small facility problems. Perhaps your system will reside in a larger facility that is experiencing hotspots or cooling flow issues. Aspen can help you with your large facility cooling issues too.
- Once installed, and equipped with adequate cooling and clean power, your cluster must integrate with your local site in many ways that may not be intuitively obvious at first glance. Does your site have a centralized user authentication schema? If so, your cluster should adhere to those standards. How about Windows integration, do your end-users work from Windows workstations? Perhaps sharing data drives to the Windows systems with Samba would be appropriate, or remote display capabilities such as Windows based X servers, VNC or ABC are needed. What about local site timing sources, IP addressing and routing for the external interfaces on your system, Internet security requirements, or even configuring your cluster for mail delivery? All of these factors and more have to be outlined, decided, and eventually configured for your cluster to successfully function as an integral part of your computing environment.
