Outsourcing to a high performance computing service provider is a good idea

Strong research and development programs are vital to the continued success of pharmaceutical and biotechnology companies in today's fiercely competitive markets. The costs associated with these R&D projects can be staggering. Also, the number of possible directions that a company can take it’s R&D program is growing at unprecedented rates. That is why many companies today are looking to reduce costs and limit their focus by turning to computer modeling, simulations, data mining, and other computer based methods. These strategies, however, are limited by the availability and effectiveness of the computational resources at hand.

Today, most companies that engage in computationally intensive research purchase their own computer hardware. Unfortunately, there are many additional costs that aren't always foreseen when choosing an in-house solution. Beyond the cost of the hardware itself, a large computer cluster consumes vast amounts of electrical power and releases most of that energy in the form of heat. Additional cooling facilities must be installed to prevent the grid from over heating. Also, a large computer cluster (and it's associated refrigeration units), take up valuable floor space. Also, costly personnel must be on staff to maintain, administrate, trouble-shoot, as well as provide end-user support.

But there are other costs that most companies probably don't even think about. For example, a computer cluster has a finite lifetime during which it can be used to perform the valuable computational research needed by its owner. However, it is rare that a company's R&D staff is able (or have the need) to utilize the cluster 100% of the time. There are many hours each day that the cluster sits idle and consumes costly resources with zero productivity. Also, as a cluster ages and more advanced computer hardware becomes available, usually at a lower price, a computer cluster doesn't not perform at the level that a similarly priced cluster purchased today by a competitor would perform. This performance gap increases in time and represents wasted CPU-time and money. This performance gap can be critically important in a fast-paced competitive research and development environment.

So what is the alternative to purchasing an expensive, resource-consuming, quickly depreciating computer cluster or a grid of desktop PCs? Outsourcing. Outsourcing your company's computational resources is an attractive solution to the costly problems associated with owning and supporting a large computer cluster. It represents a shift toward flexibility in the way companies utilize high performance computing environments. The main difference is that with outsourcing,
computational needs are met by the purchase of CPU-hours rather than CPU's. This provides research and development planners the privilege of essentially a "pay as you go" plan. Gone are the days of making appeals to the management for the large amounts of capital needed to purchase a in-house cluster, usually with very little evidence that it will be beneficial in the long run. Instead, only small amounts of money are needed to start a new study or research project.

Why does outsourcing make sense? An HPC (high performance computing) service provider is able to leverage the idle of its customers by appropriately distributing load throughout the day. The results are maximized utilization of the computer clusters and the associated savings are then passed on to the customer.

Outsourcing has many benefits over purchasing or leasing hardware. First, and most obvious, is that initial costs are reduced dramatically. Also, instead of a lengthy period of time for delivery, installation, and setup, by outsourcing, a company can literally begin their research in a matter of hours. By treating CPU-Hours as a consumable commodity, budgeting and planning become much easier and flexible. There are, however, several considerations that a company engaging in computational R&D needs to carefully examine before choosing an HPC service provider.

Tsunamic Technologies Inc. (www.clusterondemand.com) is an example of an HPC service provider that is successfully addressing the needs and concerns of its customers. TTI is a small company that began providing on-line access to its high performance computer clusters to customers world-wide in early 2000, being one of the first companies to do so. Many customers enjoy the hassle-free and personal customer service that a small company best provides. The flexibility, adaptability, and experienced personal "customer-first" attention that TTI provides gives HPC consumers value-added service that is not available with traditional in-house solutions.

By outsourcing its HPC needs, a company not only gets access to the hardware it requires to do its research, but also gains the experience and knowledge-base of its HPC service provider. For example, many of TTI's customers are relatively new to the world of cluster and grid
computing. Such customers often are not aware of simple strategies that can improve efficiency and optimize their utilization of a large computer cluster. In other cases, TTI has been able to suggest to its customers more advanced and technically challenging solutions to their unique problems. Whether it's offering non-traditional file systems, customized parallel execution environments, or any number of other specialized services, an experienced HPC provider should be able to surpass the expectations of its customers.

The primary benefit of grid computing, namely the ability to quickly and cheaply harness the computing power of a large number of CPU's, requires the integration of a grid middle-ware. This usually introduces additional constraints and complications that limit their effectiveness. Grid computing sometimes entails sending computational units, or jobs, out to various off-site facilities which are not directly accessible to the end-user. This makes debugging and software development almost impossible. By outsourcing all of its computational needs, a company doesn't have to worry about the difficult and restrictive process of integrating their resources with an nebulous grid computing environment. All of their needs are met in one off-site location, over which they have direct control.

Another benefit to using an HPC service provider instead of spending the extra effort to house your own cluster is that you are not locked into a single architecture for the life of your cluster. By using an HPC provider such as TTI, you have access to all of their varied and customizable cluster configurations. For example, some companies find themselves forced to use a legacy code that is only supported on an out-dated operating system while at the same time needing the current release of the OS to use contemporary software tools. Outsourcing can provide these companies with access to both architectures without having the addition costs of operating two distinct computing environments on an in-house cluster. Additionally, an experienced HPC provider can quickly provide customized environments personally tailored to best achieve the computational goals of any R&D program.

Purchasing computers can be time consuming. It can also be difficult to judge exactly how many CPUs your are going to need and which type of CPU will be best for your application. Not to mention the maintenance and administration costs. Most organizations that find themselves needing more CPU power tend to buy either too much computer, or not enough. It can be very difficult to accurately forecast your needs. Also, there are many factors that make owning your own large cluster impractical. Maybe you only need 100 CPU's for a few hours each day, and the rest of the day, your new cluster is sitting idle as its value depreciates exponentially. Or maybe you need 50 CPU's for 3 months to finish work for a particular customer, and won't need them after the contract is fulfilled. But usually, a company first starting an R&D project has no idea of how much computer power they will need. Outsourcing addresses all of these problems.

In general, there are many costs and headaches associated with purchasing and owning a large computing environment. Planning for the future and adapting to changing requirements are very difficult at best with an in-house solution. Outsourcing to a reliable, customer-service oriented HPC service provider is an economical and beneficial alternative. If you want to immediate access to a continuously upgraded high perform computing cluster, then a fast and flexible provider such as TTI is right for you. The greatest benefit to using an HPC service provider is the flexibility to use what want when you need at an inexpensive price.

No contact information available. Please use link to company website.