Nine Design, Implementation Factors for Modern Enterprise Storage

Enterprises that want to stay ahead of the curve now require data storage that delivers high performance, scalability and always-on availability in order to keep up with increasing data-movement speed demands. Not only that, in the current hyper-connected, data-driven world, organizations must overcome the challenges associated with real-time access to data to support business-critical workloads—all while maintaining the agility required to adjust to changing business directives. In this eWEEK slideshow, using industry information from Infinidat CTO Brian Carmody, we present nine principles designed to govern the next generation of storage. Infinidat was founded in 2011 by storage industry legend Moshe Yanai, who created the software for Symmetrix, cornerstone of the EMC storage franchise for more than 30 years.

Most large enterprises are well on their way to adopting a more agile, systematic approach to software development and deployment, and many are also embracing DevOps principles on a wide scale. There should be no tolerance for storage not to be agile as well. Business processes today move many times faster they did a few years ago, and storage can and must keep up.

Fast-moving environments require an increasingly on-demand infrastructure, particularly in test/dev and multi-tenant arenas. When your storage is API-driven, you can more easily control all elements of configuration, provisioning, maintenance and monitoring. API-driven storage makes the move from test/dev into production that much easier, more repeatable and highly reliable.

One way to boost the efficiency of your enterprise data storage system is by determining and acting upon the value of your data. Part of this value determination process is to optimize storage by understanding the relevance of the data to users’ needs and preferences. For example, which workloads require real-time access to streaming data versus access to historical or time series data for analytics? The rest of the IT stack is becoming more adaptive and flexible; shouldn’t the storage system be doing the same?

Let’s face it: All-flash arrays are great for some workloads, delivering high performance, good availability and high power efficiency. But they can get expensive very quickly as working sets start to scale beyond 100TB or so. As the volume of data continues to mushroom, and the need to derive actionable business insight from that data becomes more imperative, enterprises need a cost-effective solution to their data challenges. A smart, fast and flexible software-defined approach that intelligently combines flash with ultra-fast DRAM and spinning disks for cost containment—and great performance—at petabyte scale will be the best choice.

Next-generation storage solutions will need to meet—and in many cases exceed—the levels of reliability set by existing enterprise-class storage systems to make large volumes of dynamic data highly available. Many contemporary all-flash-array solutions, which are aimed primarily at performance, have yet to achieve levels of sustained reliability that would classify them as truly enterprise ready. Reliability in arrays built upon legacy controller architectures (as most AFAs are) requires large numbers of discrete, independent drives in a RAID (redundant array of independent disks) configuration to ensure data protection. While spinning disks are slower than DRAM and NAND, the technological breakthroughs of superior flash-optimized hybrid arrays circumvent that slowness to deliver better than AFA performance at much lower cost and much higher areal density.

Whatever modern storage system you choose, it’s going to be based on software. That software needs to operate seamlessly with, and be optimized for, whatever storage media the environment is using, now and into the future. Whether it’s the increasingly high-capacity drives now available, JBODs, flash—or even tape in some instances—next-generation storage will have to play nice with all types of storage media (perhaps even DNA someday).

We’d like to think storage is simple. But anyone who really knows storage knows that enterprise-class data storage can be anything but simple. In next-generation storage, configuring, maintaining, monitoring and provisioning storage shouldn’t demand familiarity with myriad dashboards and often-perplexing UIs. Next-generation storage will enable non-technical employees to manage every critical element of data storage on a single pane of glass with an intuitive and easy-to-use UI. For companies looking to take full advantage of cloud services, RESTful APIs will also be the logical choice for users to connect easily, interact and share data and resources.

It’s been observed that the decline in price per gigabyte and the increase in areal density of drives have both surpassed Moore’s Law. While the number of transistors on a chip may not be doubling every 18 months to two years for much longer, advances in memory technology are steaming ahead and leading to ever-greater densities in both volatile and nonvolatile media. Next-generation storage will no doubt take advantage of ultra high-density flash and hard drives to facilitate digital transformation. Disk vendors are already talking about 20TB drives on their roadmap.

Capturing, storing and analyzing data at scale presents a unique set of challenges for organizations. For any company that has a patchwork of storage solutions and protocols under one roof, there are substantial tradeoffs around reliability, access and availability of data. Next-generation storage architectures must support the goals of the organization, including providing data access and availability on demand, growing capacity and scaling performance, all while lowering both CapEx and OpEx for the CFO.