Oracle has supported the research industry for over 30 years and provides innovative cloud solutions to help advance research. After reading the blog below, take a look at Oracle in research institutions.
Jeff Covert observes where Oracle collaborates and competes with other cloud providers. Based on his observations of Oracle and other leading cloud solution providers, Jeff has written this blog to provide his perspective on the 'build versus buy' cloud approach for research. Please add your comments and feedback, I know Jeff would welcome your thoughts.
Cloud computing promised information as a utility. Data with the ease of a light switch. The electricity’s path, the circuit breakers and bulb is immaterial. You need light. You pay for what you use. IT is another utility that just needs to work. Your research sits on top of the platform. Creating the platform is a waste of precious resources.
How much time and budget was spent on wiring a cloud together that was supposed to be a utility? Did your facility and admin (F&A) expense ratio hurt your chances to win a grant? Wining research grants and funds is hard work. F&A costs impact grant opportunities. The process and evaluation issues with these expenses was engaged in “Knee-capping excellence” an article by Ronald Daniels, President of Johns Hopkins University. Dr. Daniels published his observations in Issues in Science and Technology magazine. Reducing F&A expenses is the path to more grants and more effective grant use.
Let’s look at how to limit your non-research IT expenses. There are two common types of cloud platforms and three cost saving approaches.
The two kinds of cloud platforms are: Build-it-yourself or a commercial cloud service. To build an IT platform you need to set and manage partitions, buffers, queues, and data event handling. Quick start code, an orchestration tool, and open source code require modifications. Typically the build-it-yourself labor, security, integration, and ongoing platform maintenance expenses are not charged to the project. The alternative is a low code, configurable commercial cloud service. The commercial service is faster, complete, tested, and maintained by a vendor. The build-it-yourself tool kit is cheap to start and good for small projects. Some projects start small and consume large budgets over time. Only the bravest projects track all the labor spent on do-it-yourself IT work. Frequently, this labor is buried as indirect costs. These expenses are visible in the Institution’s cost profile making new proposals unaffordable.
How can the build-your-own cloud be managed? The following cost savings approaches are tied to vendors. One approach is to preempt certain workloads during peak usage hours. The supply and demand approach will serve those that can operate their workloads during off-peak times of day and shorter duration. This approach requires that compute tasks can be stopped and restarted in lower cost time of day. Another approach is a loyalty rebate. As you consume monthly resources a cash-back award is earned for large time blocks of usage. The third approach is to have a cloud service that has AI/ML efficiencies built into the service. This is “all-in” pay as you go model does not use restarts or cash back. Sara Jeanes wrote an article in Internet 2 magazine about this topic. Sara included usage and price data for AWS and Google. A useful performance comparison between AWS and Oracle is here. An array of industry benchmarks showed the performance of the pre-engineered approach over build-it-yourself.
Research requires IT capabilities. Efficiency gives your institution an advantage in competing for funds. Utilizing a pre-built cloud platform saves costs and enables more research.
Thanks Jeff, great insight! If this resonates with you, please join us at Oracle OpenWorld, 16-19 September. I am really looking forward to talking to a great panel in the ‘Advancing Research with Oracle Cloud’ session. This and other research sessions can be found in the education and research sessions list.