Advantages of cloud-based predictive analytics




Like most industries, financial institutions depend on fast and reliable access to data. In the area of fraud, for example, a bank will often monitor a customer's account balance, spending patterns, even the way they hold their credit card - all at the same time.

If data delivery in these kinds of areas is disrupted, business will slow and customers will invariably suffer. Therefore, understanding why disruptions occur is critically important. But it is often misunderstood, says Gavin Cohen, vice-president of product and solutions marketing at Nimble Storage, the most recently acquired company of Hewlett Packard Enterprise.

"When there is a problem in IT infrastructure, storage is almost always the first thing to get blamed; but, most of the time, that's incorrect," says Cohen.

According to recent research by the company, storage-related problems are responsible for less than 50% of data disruptions. 54% are down to configuration, interoperability and other non-storage-related issues.

What we have found is that disruptions are usually caused by the complexity of the overall infrastructure. The more complex the environment - such as in a bank - the more likely you are to have these slowdowns.

"What we have found is that disruptions are usually caused by the complexity of the overall infrastructure," he continues. "The more complex the environment - such as in a bank - the more likely you are to have these slowdowns. You have servers, networks, storage, hypervisors, operating systems and applications. These all have to be working very well together to get end-to-end data successfully delivered at speed to your applications."

Pre-emptive strike

While flash can remove most storage performance constraints, flash alone is not sufficient. What Nimble Storage has created, Cohen says, is a way of monitoring a customer's infrastructure that can predict and therefore prevent data disruption.

"We have taken the unique approach of using predictive analytics," he explains. "We collect millions and millions of data points each and every second. Then we perform techniques like machine learning, and statistical and predictive analytics with the goal of finding anomalies, preventing problems and making the whole infrastructure run much better."

The results speak for themselves. According to Cohen, in a Nimble Storage environment, 86% of cases that would normally involve a customer identifying a problem and calling a vendor support help desk are found and resolved before the customer is even aware of the issue.

"That can include rudimentary issues; for example, a hardware problem where power supply voltage is fluctuating," he explains. "In this case, we can recognise that before it fails, and then trigger the process of sending a new power supply and engineer on site. Or it might be something very complex such as finding that there is a change in performance between the host and the storage. We might determine the root cause isn't the storage at all; it's down to a configuration setting in the hypervisor, for example, or in the network layer. We can then tell the customer how to resolve the problem."

By quickly spotting issues before they occur, Cohen says Nimble Storage has managed to achieve a measured availability across its entire installed base of more than 99.9999%.

"That's more than six nines of measured availability across every version of our hardware, deployed in every type of customer configuration you can imagine," he remarks. "It is really unheard of in the world of IT."

A bank's needs

Dan Young, head of IT operations and support at the independent investment house Killik & Co, says that "the reliability of Nimble Storage" has been "like having another employee on the team".

"It is important that our investment managers within the firm get their data in a timely manner so that they have the information in hand in order to provide clients with the level of service they expect," Young said in a previous interview with Future Banking. "The key issues that Nimble Storage solved for us were around ease of administration, performance and scalability."

Another challenge facing banks is moving their data into a cloud environment. This is particularly true when it comes to more traditional applications that place far higher demands on storage.

"Data has gravity," Cohen explains. "You can move an app from one place to another but, if the data doesn't go with it, the application won't continue the way it did previously. That is a challenge. If you think about large banks - each with its own specific way of dealing with data - that have petabytes of information, it is not easy to just move it somewhere else.

"If you start from scratch and go to the cloud, then you obviously have the luxury of developing applications to use all of the native cloud services. But if you are a bank and you have thousands of applications, the last thing you want to do is have to rewrite every application to work successfully in the cloud."

If you start from scratch and go to the cloud, then you obviously have the luxury of developing applications to use all of the native cloud services. But if you are a bank and you have thousands of applications, the last thing you want to do is have to rewrite every application to work successfully in the cloud.

To facilitate cloud migration, Nimble Storage has created a product that allows customers to run applications in its own system as well as move applications and data between on-premise and cloud infrastructure in a truly "hybrid IT environment".

"The platform is called Nimble Cloud Volumes, and it is an enterprise-grade block storage service for Amazon Web Services and Microsoft Azure," Cohen explains. "We have built our own clouds for storage, but we have located them in data centres at very close proximity to the public cloud. We then allow you to attach our storage volumes directly to cloud compute."

With all of Nimble Storage's solutions, security is key. "For a bank, confidence and security are everything," says Cohen. "You would never do something if it meant not being able to secure the data. What you want is to make sure data at rest and in flight are encrypted. Our approach is that all our products have encryptions. It is a no-cost feature and it is also granular. That means we can turn encryption on and off by data set, which gives our customers the ability not just to encrypt data but to securely shred it as well."

Stellar customer experience

Customer service is central to Nimble Storage. Using predictive analytics has allowed the company to completely transform its support process.

"Most support organisations are very top-heavy," Cohen explains. "There are lots of level-one and level-two support engineers who take support calls, ask very rudimentary questions, collect information and then maybe escalate to level two. You go through this whole series of tedious interactions with support that often concludes with them saying 'it isn't our problem; it's vendor X's problem. Speak to them'.

"With Nimble Storage, all of the mundane aspects of support have been automated through our analytics engine, called InfoSight. As our customer and installed base grows, that has allowed us to build a support organisation made up entirely of level-three engineers."

The company also makes sure these engineers are available quickly. "You can speak to a level three engineer in under a minute," he says. "Just contrast that with the hold time you have to suffer with a call centre. A typical vendor will take hours, days, even weeks to solve complex problems. With Nimble Storage, they are typically solved in minutes."

This customer service philosophy extends throughout the company. "We have tools in place that automate end-to-end from the very first pre-sales customer interactions all the way through the life cycle of the product installed at the customer's site," Cohen states. "In a world where it can take weeks or months for your order with a large storage vendor to arrive, it take us under a week. We have very mature processes around how we interact with our partners."

Put together, these solutions and processes have resulted in a net promotor score (NPS) - a measure used by companies to assess how well they look after customers - of 85.
"The average NPS for big storage vendors is typically in the teens to low 20s," Cohen remarks. "Our score of 85 is the highest in the storage industry and living proof that what we do is extremely effective."