Reining In Brown Energy Usage
Powering IT facilities, if done the green way, happens through renewable resources like solar and wind energy.
When renewable energy gets into the IT power mix, however, it adds a new wrinkle. While data centers and facilities could simply treat that power the same as that of coal or oil, regulating workflow such that the majority of IT functions run on green energy would be a significantly more environmentally friendly way to operate.
While that should not seem like much of an issue, power is power no matter how you acquire it, the fact is that renewable energy sources such as wind and solar are obviously intermittent. It is not always windy or sunny, and oil and gas reserves, wither produced in-house or more likely purchased from the local power company, fill in the gaps.
The challenge then lies in allocating the more energy-intensive IT work to times when renewable energy is powering the data centers in question. Research out of the Nanyang Technological University in Singapore looks at a database system called Renewable Energy Integrated Database, or ReinDB, that is designed to smartly schedule workloads such that they align more with renewable energy needs.
“The key challenge of exploiting green energy sources is that they are variable and intermittent,” noted the mostly Singapore-based team to introduce their ReinDB analysis. “So far, there has been little work on integrating renewable energy into a database system.” It is the function of the database systems to allocate and schedule jobs, a task that is usually optimized for performance. That paradigm does not have to necessarily change in scheduling the brunt of the IT work in environmentally friendly conditions.
The researchers distinguished the two types of energy as green and brown, a distinction that should be fairly self-explanatory. “The design goal of ReinDB is to minimize the brown energy consumption on a database server with both green and brown energy supplies. Specifically, we develop the green supply driven execution paradigm and adaptive power management techniques to adapt to green energy supply.”
ReinDB allocates usage through something called supply-driven execution (SDE), which, as its name suggests, adds an element of tracking renewable resources in order to schedule, or execute, jobs. “ReinDB runs on a database server powered by both green and brown energy sources. Workloads are submitted with their predefined slack times. The slack time allows us to adopt the green supply driven executionparadigm.” Knowing a workload’s slack time is an ordeal in itself before a scheduler has to consider an environmental factor in scheduling that workload.
By utilizing the SDE in concert with a performance scheduler, that task becomes manageable. As the research team puts it: “We implement the supply-driven execution on a database server with power management (voltage/frequency scaling on CPU) and workload scheduling to avoid slack violations.”
Sometimes, however, when such changes are made to the typical database scheduling system, more bottlenecks than usual appear. ReinDB reportedly deals with that by using a multiqueue algorithm. According to the research team, “In order to resolve the aggregated performance bottleneck, we leverage workload prediction and develop a multiqueue algorithm to schedule the queries according to the processing capability.”
While further technical detail can be found in their paper, it is important to note how successful these systems were in increasing a test case’s environmental efficiency.
Early returns show pretty well, as the database system accomplished a 57 percent reduction relative to “green-oblivious” systems in brown energy usage, and a 53 percent reduction against “basic greenaware” systems.
The figure below represents data taken from a simulation on Google’s hypothetical workload.
As one can see, the usage of green energy roughly lines up with that of the IT workload. While the correlation is less than perfect, it is a good start for the kin of a system that has yet to be developed but shall be essential for the continued growth of green IT and IT in general.
Whether ReinDB takes the reins on green database management systems or some other system props up to take its place, the researchers are optimistic about the future of optimizing hybrid green and brown IT facilities. “We expect,” they said in conclusion, “to see new green data management systems with renewable energy at larger scales (such as clusters and data centers) emerging in the future.”
Indeed, as the months go on, green-focused database systems should become more commonplace. Advancing the energy efficiency and environmental friendliness of IT equipment should be just as important as developing new cooling techniques and technologies for green computing to progress.
Related Articles
Heat-Trapping and NREL's Green Datacenter Leadership Effort
Does Green Sacrifice Performance for Efficiency?
eBay Builds Energy Efficiency Dashboard
