The Chicago Cubs are now the 2016 world champions of baseball. I’m not a Cubs fan, but I couldn’t help rooting for them given that the last time they won a World Series was back in 1908. For you trivia buffs, in 1908, 90% of all households had no electricity, the average wage was 22 cents/hour, and average life expectancy was 47 years. It was also the year that the first Model-T rolled off the assembly line, making automobiles more affordable, and over time making Henry Ford a very wealthy man. At that time, only ~80,000 cars existed in the US with only a few hundred miles of roadway suitable for automobile travel.

OK, so what does all this have to do with hyperconvergence you ask. (We’ll get there, I promise.)

Well, the assembly line process for building cars radically changed the transportation industry. It made automobiles plentiful and put them within reach of the masses. But, the practical adoption and massive proliferation of automobiles in society would not begin to occur until 1916, when government laws and money began to fuel the building of “surfaced” roadways. By 1935, more than one-third of rural roads in the country were surfaced, mostly paved with asphalt or concrete. The muddy roads connecting the country’s growing urban centers, most of which were impassable for much of the year, were vanishing, and for the automobile, as they say, the rest is history.

It’s now 2016 and Hyperconverged Integrated Systems (HCIS) are the next great “thing” in IT circles, promising a more efficient, agile, and cost-effective building block approach for modernizing the enterprise datacenter and cloud. Industry pundits forecast explosive growth for HCIS over the next decade, and all of the major IT vendors are jumping on the hyperconverged bandwagon for a piece of the action.

But like the automobile and road situation in 1908, hyperconvergence will not “take off” at scale until those muddy roads that connect the clustered nodes that comprise HCIS are fixed. Sound’s mundane and trite, but the “fat dumb pipes” that make up the traditional network infrastructure, which by the way hasn’t fundamentally changed in more than 25 years, simply doesn’t fit the bill for this next wave of transportation, oops, I mean Information Technology.

Recently, Gartner published an excellent report, entitled, “Leverage Networking to Ensure Your Hyperconverged Integrated Systems Can Support Demanding Workloads” that puts a spotlight on the network challenge of HCIS. (You can access this report here).  Anyone who plans to build, sell, or consume HCIS should read this report. If they don’t, they’re ignoring a very important and fundamental piece of the HCIS puzzle.

As Gartner points out in their report, “Most HCIS deployments today have largely ignored the network. Adoption of HCIS has built a new silo, resulting in new coordination challenges for the network team. Mixing user access, internode application traffic, internode storage traffic, VM mobility, and storage access traffic on a single network switch without bandwidth control can lead to unpredictable performance and system stability, and compromised data integrity.”

Basically, what Gartner is saying is that HCIS introduces new network workloads that are placing new requirements on the network, requirements that traditional networks were not designed to address. Most early HCIS deployments consist of a handful of nodes “often hosting less-demanding workloads such as virtual desktop infrastructure (VDI), application development, and remote office/branch office infrastructure.”  It’s when these HCIS clusters grow and begin to support applications with more demanding I/O workloads that network traffic challenges become real problems, unless the network component is addressed. In fact, according to Gartner, the I/O loads alone “will increase by 10x” with HCIS clusters.

Multiple, mixed workloads, significantly increased I/O traffic, and deployment at scale demands a network that can isolate and control HCIS workloads to ensure performance, security, and availability. Here, control is key. Without this level of granular control, workload traffic will become congested, and scale will be severely limited.

Plexxi’s Infrastructure Network Fabric can provide the level of control and workload isolation crucial to HCIS. With Plexxi, organizations can deploy a single IP network that supports all HCIS and application workloads. Network administrators can define workload policies, allocate path bandwidth, and isolate workload traffic across the Plexxi fabric, eliminating the need to deploy separate IP networks to meet performance, availability, and security SLAs. As a software-defined network architecture, Plexxi provides the management interfaces and application APIs needed to build an agile datacenter network.

If you get a chance, read the Gartner report and our recent newsletter posted here, especially if you’ve deployed or plan to deploy HCIS. It will help you understand the fundamental role networking plays, and you’ll see why Plexxi is a perfect match for HCIS. We’re not fat, dumb pipes; we’re intelligent networking under software control. So, don’t get stuck with muddy roads. Let Plexxi help pave the way to your HCIS future.

Leave a Comment