The word “innovation” gets thrown around a lot these days. I wouldn’t blame anyone if the word now evokes a gag reflex similar to how we all eventually reacted to “cloud” a few years back. Every product and presentation seemed to be reworked to make it appear it was a “cloud” product. It made sense since manufacturers were chasing the cloud’s total addressable market (TAM) that folks like Gartner and others had so accurately forecasted, but nevertheless it made for some confusing times as buyers were forced to wade through the misleading marketing to see what the reality was. As an industry, we spent a lot of time simply trying to get organizations to all agree on a common definition for cloud like the good people at NIST and Wikibon provided. Once we were able to establish that common language only then could we talk about what solutions were available to solve the challenges that business faced. The reality was people were actually looking for some pretty basic things. The market wanted solutions that were reliable, easy to use, secure, flexible, scalable and cost effective. We can talk about deployment models (SaaS, IaaS, PaaS) but it all started with the first set of requirements. As we now move beyond conversations about where (aka private, public, hybrid), we seem to have landed at a consensus that the world is now multi-cloud. In this multi-cloud world those requirements are applied regardless of “where.” Whether it is here or there (starting to feel like Dr. Seuss) customers expect that solutions will meet the attributes I described above.
The web scale companies have done a great job setting the bar by making infrastructure invisible. I have jokingly referred to it as the “IT vs. Amazon” phenomenon. Deploying applications on Amazon is amazingly simple. Compare that to how we traditionally deploy applications on premise and it’s a knockout win for team AWS. As an industry we continue to make strides in improving this user experience but we still have a long way to go. Hyper-converged infrastructure innovations have helped narrow this gap in user experience tremendously, but we still have work to do. If we can all agree that the basic underpinnings of any infrastructure stack are applications, networking, compute and storage, then we can see why the HCI folks can’t solve this problem alone. Before people start lighting torches and picking up pitchforks, I’m including orchestration software under applications. I agree it is critically important to delivering the true end user experience we are after, I’m just not going to focus on it here.
We have seen lots of innovation (there’s that word again) in the application space with containers, microservices and scale out designs. The compute and storage tiers have been impacted by virtualization and/or abstraction, and have seen increased performance via SSDs and emerging NVMe, as well as the physical convergence into one form factor, and the HCI management frameworks now also include orchestration. I could write an entire post on this topic alone as there are many advancements that have radically changed the way we do things. The one outlier is this solution stack is the network.
The folks at Gartner recently published a research note Look Beyond Network Vendors for Network Innovation. There is that “innovation” word again. In that note, they talk about how innovation isn’t coming from the traditional network manufactures. They talk about how those same vendors overhype even minor feature enhancements as “innovative”, and how true innovation is being created and driven into the market by the web scale (hyperscale) companies. That note was the inspiration for this blog post and is why I believe the industry has hijacked the word innovation much like it previously did with cloud.
Gartner’s position shouldn’t come as a huge surprise. This premise has been discussed at length in books such as The Innovator’s Dilemma. Big companies typically struggle with true innovation because they come at it from the lens of their product set, and the need to protect existing revenue streams. This leads to more iterative approaches that still move technology forward but albeit at a slower pace than truly disruptive technologies. The big companies are also typically slow to react to emerging markets because they are after larger market share opportunities. These points and others in the book illustrate why big companies don’t typically innovate. Instead, they iterate. Now, don’t get me wrong, iteration is not bad. But there is a difference between an innovation that creates new ways of doing things, and an iteration that simply does the same thing differently and slightly better. While the compute and storage side of the data center have seen innovation, I would argue the network side has seen mostly iteration.
Folks like James Hamilton from Amazon have railed against the network, saying the datacenter network is in his way. To that point, the web scale providers realized that they couldn’t continue building networks the traditional way. It would be too expensive, too complex, and too difficult to manage. So, they set out to build a better network. The basic principles of the design would be one system – built on commodity switches, 100 percent software defined and with deep API integration, that would allow any workload to be deployed anywhere, at any time. So off they went building their networks and at the same time creating an optimized supply chain for low cost, high performing, commodity off the shelf (COTS) switches.
In my previous stint at a national solutions provider, we met with lots of customers who wanted to be like the web scale providers. The problem was most companies don’t have teams of developers to build it, the resources to support it, and more importantly the network wasn’t really a key part of their core business plan. What do I mean by that? If you are in the business of making shoes, you probably don’t care to make a network too. It’s a waste of precious capital and time. For the web scale providers, the network is essentially part of their product. So, while customers wanted to be like the web scale companies, it wasn’t really feasible. Until now.That’s where Plexxi comes in. The founders here at Plexxi started with a vision of building a better network. They wanted to provide lots of diversity without complexity. They wanted to build that network using the same principles of the web scale providers, one system, built on commodity switches, 100 percent software defined, with deep API integration, that would allow any workload to be deployed anywhere, and at any time. Leveraging the supply chain that the web scale providers created, Plexxi has indeed built a better network. Google liked it so much, they invested in us. Plexxi built a network that aims to put the software defined back into SDN. We do that with our application and workload integrations. These integrations can pull and push data back and forth, understanding intent and creating actions based on that intent. Did you know that the number one support issue for HCI is network configuration? We solve that by automating the VM lifecycle. We discover the VM, adapters, interfaces and LAGs, we auto-configure LLDP, CDP andVLANs and we segregate storage traffic eliminating the need for separate SAN fabrics. We provide end-to-end visibility for both the VM/cloud and network administrators. For the VM/cloud admin, that means no more opening a trouble ticket with the network team when deploying new VMs. For the network admin, that means not having to be pulled away from higher level business value projects to handle trivial deployment tasks. Everyone wins.
Now while the HCI use case for Plexxi is extremely powerful, it isn’t all we do. With all of our bandwidth and diversity we are also a better network for data intensive environments (HPC, Hadoop) where traditional oversubscribed architectures cause performance issues. We can isolate backup traffic to avoid contention, and we have integrations with VSAN, AHV, NSX, OpenStack (with more coming)! Heck, we are a better network for almost anything. The beauty is we can start small with a couple of switches, and scale big supporting companies like Perseus where together we won an innovation award for building the largest software defined network in the world. Now does that sound like iteration or innovation?
So, listen to Gartner, and give us a look as we bring a productized version of web scale networking to the market. VMware didn’t invent virtualization, they productized it and made it consumable for the masses. We are doing the same thing in networking. This is going to be a great year for Plexxi as we continue to help our customers build reliable, easy to use, secure, flexible, scalable and cost-effective cloud infrastructures. Give us a look, you won’t regret it.