SDN is the Next Great Hyperconvergence Frontier
[dropcaps type='square' color='#ffffff' background_color='#e04646' border_color='']O[/dropcaps]ften overlooked, networking is a vital component in a hyperconverged. One reason it is often overlooked is many customers chose to use an existing 10 Gigabit Ethernet environment, and they simply plug their hyperconverged infrastructure in and go on their way. Other customers may choose to put in a brand new 10 Gigabit Ethernet network, and begin their hyperconverged environment with a fresh base.
Many hyperconverged solutions have two 10 Gigabit Ethernet ports for connectivity – and that’s it. Some solutions have the ability to expand beyond two 10 Gigabit Ethernet ports to address various additional high performance needs. They may also have Gigabit Ethernet ports to provide connectivity in small branch offices that do not have or do not need a 10 Gigabit Ethernet infrastructure.
This is a big change from the ESXi hosts administrators are used to. Many initially are skeptical of running their host with just two network ports, no matter how fast they are. Many administrators are used to having confusing network configurations on their hosts along with the pain of managing and modifying them. Hyperconverged infrastructures are designed with simplicity in mind, and the network simplification is a huge part of that. Gone are the days of hosts with 12 network cards, and complex configuration on network switches.
The Aggregation Trend
With higher speed networking comes the reduction in overall port count. This is largely regarded as a positive selling point. At the same time that we have won on the network footprint of our hyperconverged platform, we are now faced with the challenge of providing the best strategy to distribute traffic through the data center.
There are some conversations around the concern of increasing the density of workload on a smaller port count, but as most operations teams will discover with monitoring of their environment, the networking is almost never a constraint. The topologies are changing as more networking teams build the desired spine-leaf architecture where core switches aggregate the uplinks to leaf switches in the top of rack (ToR) for the compute platforms. This is the North-South networking traffic that has been traditionally seen in the data center.
As the data center traffic grows with increased application workloads, we are seeing a rise in the East-West networking traffic. East-West traffic is the VM to VM traffic that is increasing as more development is happening around N-Tier application architectures, as well as the density of workloads in the compute platform. More workloads are able to communicate over the lowest latency connections in the same compute host.
Some view this as a potential conundrum as it creates a potential point of failure by increasing the density of workloads on a hardware platform. This is answered by the resilient physical architecture of hyperconverged infrastructure. Every point of operation in the hyperconverged architecture is designed to withstand failure without impacting operations. There will always be a level of failure that could occur which could cause an interruption or outage, but the redundancy and resiliency of the hardware provides a robust level of protection. Operational software and the hypervisor become the next layer where we move to create resiliency.
As the hypervisor becomes more tied in with the virtual and physical networking, we have a lot of evolution happening in design and operation of networking platforms thanks to hyperconverged architecture.
Built for SDN
The core of hyperconverged infrastructure is the software. Simplivity and Nutanix have both shown that the heart of their companies is the software that drives the platform. While they offer a hardware delivery system that leverages the software they develop, the real value comes with the API-driven, software platform that allows for feature growth and evolution of the platforms.
Not only has the compute layer been designed for a software-defined approach and all the abstraction to create better management and operational capabilities, but the storage has been designed the same way. Beyond the compute and storage, we have one remaining key layer to be pushed to the next level: networking.
Software-defined networking (SDN) is gaining in awareness and popularity, with good reason. VMware is putting incredible weight behind the engineering and marketing of their NSX product. Cisco is doing they same with their ACI ecosystem. Disruptors are coming into the marketplace with SDN options for existing network infrastructure as well as some hardware offerings as well. The shift is underway for many parts of the industry, and as the virtualized data center is evolving with hyperconvergence, it is ideally designed to leverage SDN architecture for the new mode of operations.
Extending SDN into a hyperconverged architecture could be one of the simplest of all of the steps as we embrace the new ways of operating the data center.
Scaling for Security with SDN and Hyperconvergence
SDN provides enhanced capabilities for the data center, primarily with its ability to provide policy, and security into the networking platform along with the automation and programmability that has been sorely lacking in legacy physical networking platforms.
Hyperconverged platforms provide the same programmability, scalability, and predictability that are perfectly placed along side the new management and operations methodology with SDN. As we drive towards more orchestrated infrastructure, the natural extension is to build that orchestration framework to go end-to-end to ensure that the workloads are deployed into the storage, compute, and now the networking that is required.
The advantage with hyperconverged offerings that are in the market today is that they are built to be able to support the traditional network architecture, and also designed to integrate beautifully in the more modern SDN platforms. Just like the VDI and cloud products that are driving reference architectures for the repeatable, consistent deployment on hyperconverged platforms, we are also seeing SDN offerings that are being built to map against the available hardware. It won’t be long before we may see SDN reference architectures
So, whether you are deploying for today’s networking design, or building towards the next level of architecture in your environment, you can feel assured that the hyperconverged infrastructure is going to fit well regardless of where your organization is in the networking evolution.