“Humans have long been adept at analysing information received. Now imagine the power of a machine to breathe in data, study it, learn from it and exhale an accurate solution – that’s machine learning. It’s a neural network on steroids,” says Anton Jacobsz, Managing Director at African value-added distributor, Networks Unlimited.
He highlights that a large focus today is on how machine learning can assist businesses to better serve their customers, be it through personalisation or self-service. “However, machine learning can also be used to run your computer networks more efficiently,” he says. “In particular, algorithms can be applied to accurately predict computing requirements – from storage capacity needs to memory demands. This precise forecast leads to higher utilisation of systems.”
Jacobsz takes the example of Tintri, a brand that Networks Unlimited distributes in southern Africa, highlighting how its analytics technology combines machine learning algorithms with virtual machines and container-level granularity to predict storage and computing needs.
“Tintri’s fundamental difference is its ability to provide analytics on every virtual machine (or container) in a user’s footprint, whether those VMs are running on its latest all-flash arrays or older systems,” explains Jacobsz.
A recent blog on the Tintri site delves into the benefits of predicting storage performance, stating that useful predictions of storage performance require two things:
1. Predicting what your performance demand is likely to be in the future; and
2. Understanding the performance characteristics of the set of storage arrays in your environment.
A further excerpt reads: “We’ve [Tintri] worked to understand the available machine learning tools and what problems each one is suited for. We test those tools against real Tintri customer data to determine what works best in Tintri environments.
“In this case, our prediction method involves a linear regression of a variable-length tail segment of the data set. We use a long- and short-term predictor and merge the results of the two. A heuristics-based approach is used to decrease the time to solution.
“Knowing storage performance requirements doesn’t do much good without understanding the performance capabilities of storage systems and the performance demands of different workloads. Tintri has gone to great lengths to model these as well. From a performance standpoint, this includes understanding: How expensive is a large I/O versus a small one? How expensive is a read versus a write? How expensive is sequential versus random I/O?”
Predicting how much storage is required in a data centre is not as straightforward as it may seem, due to the ever-present complexity of compression and duplication, and a mixture of hybrid models found in data centres.
“The role and impact that machine learning can play in all areas of business should not be underestimated. It is not only an algorithm to predict human behaviour, but an excellent crystal ball into the future of your technology network too,” concludes Jacobsz.
