Corporations from many sectors have learned that grids within organizations or departments can not only increase processing capabilities, but also cut operational costs. Execution management solutions such as Sun Grid Engine or Platform LSF, as well as products from DataSynapse, United Devices or Cluster Resources, integrate with data access and data caching solutions. Together they can form an architecture capable of speeding up applications several-fold, by distributing the data and workload among the IT resources inside a company.

For mature companies looking at further virtualizing their infrastructures, a natural next step is to engage external processing resources. External cycles can prove even cheaper than the internal grids. The resource provider allows for partial outsourcing of the grid maintenance and administration, reducing both the need for staffing and project complexity.

Strategies differ when it comes to scaling up compute capabilities. A relatively standard, but not always optimal, choice is to connect to a large provider of on-demand compute cycles. Today, several providers compete in providing their virtual machines, with differing qualities of service, types of service-level agreements and prices. Another interesting choice to consider is integrating with multi-institution research projects looking for ways to promote and utilize their grids. Some of these initially offer their services at a bargain price, complete with advanced software service stacks.
 
The decision to adopt external processing solutions involves information security analysis, as business-critical data cannot travel outside corporate firewalls. Another frequently quoted barrier that needs consideration is licensing, as the license provider is required to run instances of the production software. However, the economic benefit calculation and analysis that usually drives the adoption of the external processing model often counterbalances the technical challenges. To take an example, an engineering company could shorten its production cycle by a month if simulations were run three times faster. This means hitting the market with a product one month faster, which, in turn, can be calculated in terms of tangible monetary value. This is the way in which ROI for use of grids needs to be evaluated. In a similar way, a bank needs to evaluate the profit that could be gained by knowing its risk position minutes faster than competition. A post-processing shop needs to price its potential readiness to undertake larger contracts. The same holds for a biotech startup, usually working for larger customers, which could take up larger workloads if extra processing was available. Understanding the ROI of different grid solutions is a prerequisite to the discussion of the most appropriate resource scaling strategy.

Last year, for example, a large manufacturer of electronics considered outsourcing its processing to its partner in the EGEE ecosystem, currently the world’s largest partner grid. GridwiseTech, as a vendor technology adviser, played a key role in this project, integrating the pilot program into the EGEE environment. Apart from their independent consulting services, GridwiseTech also provides a software solution that allows users to push jobs from corporate resources to the grid middleware of EGEE.

At the forthcoming OGF/EGEE User Forum conference in Manchester, I will host a seminar explaining strategies for scaling processing capabilities to cut processing costs, including connecting up to external resources. The session, “How Business and Industry Can Profit on Using the EGEE Grid,” takes place at 2:45 p.m. on May 10. It is part of the Users track and is open for conference visitors. Participants are welcome to contact the tutor in advance, using the GridwiseTech contact form (http://gridwisetech.com/contact), in order to steer the session toward their business interests.