On Dec. 3, 2015, I moderated a “Top 10 Things About SDI” panel at the Software-Defined Infrastructure (SDI) Summit in Santa Clara, CA. The panel discussion included IT company executives from IBM; Gridstore—a Silicon Valley hyperconverged systems provider; NodePrime – a San Francisco software firm; and FieldView Solutions, a New Jersey data-center monitoring and consulting firm. The panel built on the key takeaways from the SDI Summit’s keynotes and sessions.
Although SDI is still taking shape in many data centers – starting with Cloud Service Providers and Large Enterprises – it is gaining traction, because data centers have too many servers, storage devices and virtual machines to manage the old-fashioned way. Instead of manual control, using management consoles, more workload assignment is being prioritized and deployed via automated processes with SDI.
Building on Earlier Technology Waves
Why will SDI succeed? It is clear that SDI is building on earlier technology waves for virtualization and cloud, which are making it possible to move applications to the data or to computing resources – rather than the other way around. And I believe that the advent of hybrid clouds will require faster, better workload management, which is aided by the policies and automation of SDI technology. Hybrid clouds link enterprise data centers (on-premises) with public clouds (off-premises).
The SDI Summit panel has already seen many deployments take place, informing its opinions about “lessons learned” from SDI projects. The panelists included: Doug O’Flaherty, Portfolio Management of IBM Spectrum Scale in IBM’s SDI group; George Symons, CEO of Gridstore, a hyperconverged systems provider in Mountain View, CA; James Malachowski, CEO and co-founder of NodePrime, based in San Francisco, CA; and Sev Onyshkevych, CMO of FieldView Solutions, an Edison, N.J., consulting firm and provider of data center infrastructure monitoring (DCIM) software.
Top 10 Points About SDI Panel’s Key Take-Aways
The Top 10 Points about SDI that surfaced through the panel discussion are shown here. Most of these shed light on the challenges in absorbing this technology – and gaining its most effective use – in focusing on the workloads that run the business:
- Workload-Centric Optimization is Key to SDI
- SDI Should be Policy-Driven at Each Layer of Abstraction
- Visibility in SDI is Hard – It’s Important to Get to the Root Causes of Failure
- CSPs (Cloud Service Providers) are the Model [for Using SDI] – Bring the Lessons Learned “Home” to the Hybrid Cloud [linking enterprise data centers and cloud service providers’ data centers].
- Location Still Matters – Know Where Your Data and Users Are
- Scale Matters: Fit Your SDI Infrastructure to [the Places in the Network] Where you’re going to scale
- Right-Sizing: You Don’t Have to Own the Third Rack (if you already have Two Racks On-Premises, at Your Data Center)
- Experience Counts—It’s a Scarce Resource; Leverage IT Skill-Sets for Management of Workloads and Data
- Apps are Easy to Move – while Data is Not Easy to Move
- SDI has a Feedback Loop – via Analysis, Analytics and Monitoring. As IT Professionals, You’ve Got the Data to Adopt and Adapt.
What Does All of This Mean?
There’s a learning curve to climb. Take a close look at the Top 10, and you’ll see that many of the lessons learned were hard-won, and benefited from deep analysis into how it works. For example: Moving the apps is easier than moving the data. For example: Analytics can be used to identify bottlenecks, and place data efficiently.
Typically, SDI deployments are phased into a data center, starting with the first projects and building on the lessons-learned in following deployments. There is a history here, too: Cloud service providers (e.g., Amazon Web Services [AWS], Google, Facebook and Microsoft Azure) tried SDI management first – and custom-coded much of their SDI functionality.
Today, more products and toolkits are becoming available in the growing SDI marketplace – including hyperconverged systems (combining compute, software and storage); software for management, monitoring and task-scheduling, distributed data analytics software, automation software and flash-based storage.
Communicate and Network to Learn More Quickly
The effort is clearly worthwhile to those who deploy SDI – allowing them to better manage and control large numbers of servers and storage devices in large data centers. For SMB companies, with fewer IT staffers, and more reliance on outside providers, SDI is part of the “glue” that will link their own data center with cloud service providers (CSPs), managed service providers (MSPs) and hosters running some of their applications remotely.
Given the growing interest in SDI for data-center infrastructure, I believe that it’s vitally important to “network” with other IT professionals who have started these SDI deployments. If you want to make your SDI journey smoother, and safer, find these early adopters of SDI: Seek them out, meet with them –and learn from them – before planning or building your own SDI projects.