Why mission-critical storage needs more intelligence

How to predict - and prevent - errors


Sponsored Storage is perhaps the most fiendishly complex part of enterprise IT, as the storage infrastructure has to meet the demands of a range of workloads with differing requirements for performance, all while ensuring reliability. This is particularly so for storage that supports mission-critical applications where high availability is an essential requirement.

Add to the mix the need to accelerate the speed of their business and become more agile, and it is clear that organizations need to have a strategy for managing their data, one that takes a more intelligent approach. This in turn calls for more intelligent platforms to support the whole process and avoid being tied down with the complexity that mission-critical storage often entails.

These are the challenges that HPE set out to address with its Intelligent Data Platform and the HPE Primera mission-critical storage arrays that are based on it.

Although HPE Primera builds on the 3PAR heritage, HPE has taken the opportunity to effectively start with a clean sheet and build a next-generation solution that addresses customer challenges, Matthew Morrissey, senior product marketing manager for HPE Primera and 3PAR, says.

“The roots of the platform really started when we were talking with our customers and prospective customers about their specific pain points in this mission critical space. What we were finding is that a lot of these organizations were just spending too much time managing, supporting and tuning their infrastructure, and it was holding both IT and the business back, because their time was consumed with just keeping the lights on versus driving innovation.”

Part of the problem is that the basic architecture for mission-critical storage is little changed from when SAN systems were introduced around 20 years ago, yet the workloads that organizations are running today have evolved drastically. Storage arrays have gained numerous enterprise grade features, such as snapshots and remote replication, but these have arguably contributed to the increased complexity of configuring and managing the storage infrastructure.

High availability, one of the most important considerations in the high-end storage space has traditionally been achieved by building lots of redundancy into the system, so that once a failure occurs, you have another redundant piece of hardware ready to fail over to.

This reactive approach is outdated, according to HPE, and a better solution is to be able to predict errors before they occur and therefore prevent them. This implies the use of AIOps (artificial intelligence for IT operations) to use a mixture of analytics and machine learning to spot any developing issues and take remedial action.

“With the coming of intelligence and AI, a much more efficient method is to architect a system with AI in mind, where you're looking at a more predictive approach, preventing problems before they happen,” says Morrissey.

This capability is delivered through HPE InfoSight, which the firm gained via its acquisition of Nimble Storage several years ago.

InfoSight is a cloud-based service that collects telemetry data from HPE storage systems deployed at customer sites worldwide, and uses machine learning to detect patterns or signatures that could indicate developing problems. It has been operating for over a decade and accumulated trillions of data points, making it widely recognised as the most mature predictive analytics platform for IT infrastructure.

HPE has expanded InfoSight’s capabilities by embedding Primera with an embedded AI engine for a complete end-to-end AI pipeline. This means HPE Primera is globally informed and locally optimized. Bringing some intelligence onto the system itself allows Primera to operate in real-time to ensure predictable performance for workloads within the storage environment.

“Every moment of every day Primera never stops predicting future application performance and resource needs in tiny increments. Every few seconds, Primera has a continuously developing window into the future and this enables the system to intelligently and dynamically optimize resource utilization to drive predictable performance,” Morrissey explains.

With Primera and the Intelligent Data Platform, HPE is bringing the intelligence further up the application stack, to be able to pinpoint problems at the virtualization layers, such as in VMware and Hyper-V. This matters because more than 90 percent of issues affecting performance happen above the storage layer, according to HPE, and identifying and correlating those issues is a problem that’s become too complex for humans to solve.

As an example, adding a new workload to a storage array can often have unforeseen implications for the performance of other workloads. Thanks to InfoSight, HPE Primera understands the workloads and their required performance and allows customers to comprehend what will happen if a specific application were deployed on their storage array, and plan accordingly.

This includes recommendations for expanding storage capacity, or perhaps just moving workloads around to make the optimal use of available resources in the entire storage infrastructure.

“From a workload planning perspective, we can help you optimize your environment by putting a specific workload on a different resource, because maybe you don't have to buy additional hardware, maybe you don't have to buy additional flash. We can help you optimize your environment with your existing resources,” says Morrissey.

This self-managing automation and simplicity should not come at the cost of performance, and HPE Primera delivers this through a high degree of parallelization and an all-active multi-node architecture that has been built for NVMe drives and architected to deliver ultra-low latency. “When we look across the installed base of Primera, we know from the figures we get from InfoSight that 75 percent of all I/O is delivered within 250 microseconds of latency. And that just screams predictable performance,” Morrissey says.

So confident is HPE of the performance and reliability of the Primera arrays, it is backing them with a 100 percent availability guarantee as standard, with the proviso that customers are up to date with Primera OS updates. Customers can claim credits for up to 20 percent of the value of the array in the event of any outage.

The level of automation in HPE Primera extends to deployment and upgrades. While traditional mission-critical storage infrastructure often requires professional services technicians to come in and install, HPE claims that customers can deploy Primera themselves in as little as 20 minutes. Software upgrades are also seamless, with InfoSight informing customers what upgrades to download rather than an administrator having to pour through release notes. HPE claims that customers can upgrade Primera themselves during business hours, with no disruption to service.

Organizations also now live in a hybrid world, with workloads spread across on-premises and cloud-based infrastructure. There is a need to move data seamlessly between the two environments, such as to move an application from a cloud-based developer environment into an on-site production environment, where low latency and high availability can be guaranteed.

To deliver this, Primera includes native integration with HPE Cloud Volumes and with cloud providers such as AWS and Azure. The cloud management tools embedded into HPE Cloud Volumes help avoid the hidden costs such as egress fees, according to HPE.

This also touches on the issue of data management. Throughout the data lifecycle, data gets moved primary storage to secondary storage to backup and archive storage. The Intelligent Data Platform moves the data to where it needs to be based on the requirements of the application. HPE InfoSight also provides customers with the ability to forecast capacity and performance needs and provide recommendations for timely upgrades, where necessary.

With HPE’s GreenLake subscription payment model, customers can also avoid the need for a large up-front capital expenditure, and HPE can even take the entire weight off the customer’s shoulders with a team that will remotely manage Primera as well, if required.

What this all adds up to is that organizations need greater intelligence in their storage infrastructure, in order to help them meet the growing demands of data and mission-critical workloads. They need an intelligent platform that can predict and proactively resolve issues before they occur, and deliver recommendations about how to handle specific workloads. Powered by InfoSight, HPE believes that Primera and its Intelligent Data Platform delivers on those requirements. Primera raises expectation for mission-critical storage with a 100 percent availability guarantee while delivering the agility of cloud.

Sponsored by HPE


Biting the hand that feeds IT © 1998–2021