Hyperconverged infrastructure (HCI) is expected to take the enterprise by storm in the coming years, replacing today’s Goldberg-esque collections of servers, storage and networking gear with sleek, all-in-one compute modules.
But this does not mean any of these functions cease to be important to the enterprise, either in an operational/performance or service/maintenance sense. In fact, organizations will have to weigh competing attributes even more carefully given the ease with which these systems can be deployed.
This is particularly true of networking. Most hyperconverged platforms will feature internal interconnects that create mesh-style fabrics between their own resources. Individual clusters of both physical and virtual resources will require a separate layer of networking that, depending on the configuration, might require longer-form infrastructure, most likely housing any number of virtualized network architectures. And then links to the cloud or other remote sites will require long-haul connectivity, which, again, will probably incorporate SD-WAN technologies.
Clearly, there is a lot at stake for today’s networking companies. Hyperconverged vendors like Nutanix and HPE already incorporate a range of networking systems in their platforms, but as the market unfolds it seems they are in constant need of ever more flexible and efficient ways to connect disparate resources.
Nutanix, in fact, has been working with Juniper Networks for some time, most recently incorporating the QFX line of switches and Contrail Enterprise Multicloud software into the Nutanix Enterprise Cloud platform. The goal is to offer the enterprise an integrated solution that streamlines local data infrastructure even as it boosts massive scale by harnessing multiple clouds under a single, federated data framework. Meanwhile, an earlier collaboration between the two companies allows Juniper’s Prism management engine to employ broad network visibility and advanced analytics to extend management capabilities — and ultimately fully automated oversight — across distributed architectures.
For its part, HPE is adding new capabilities to the SimpliVity HCI platform. The latest version sports a composable mesh fabric based on technology it acquired from Plexxi earlier this year, which HPE execs describe as self-optimizing and capable of pushing the economics of HCI to new levels. The aim is to provide a level of infrastructure autonomy to virtually everyone in the enterprise by removing the need for IT specialists to define and provision needed resources. The same mesh technology is also going into the HPE Composable Cloud, which is now available on the ProLiant DL rack server.
Other vendors are looking at HCI on a more granular level, offering network-specific platforms that can be used to support low-cost private cloud infrastructure. Array Networks recently took the wraps off the newest AVX series appliances that deliver network functions capabilities, KVM container support and security to local scale-out environments. The system runs on Array’s network OS, while also supporting third-party solutions from Cisco, Palo Alto, F5 and others. It can also be configured with 40 G Ethernet.
Even in the modular, hyperconverged world, networking is still a complex task. Fortunately, with new levels of intelligence and automation, much of that complexity can be masked under layers of abstraction.
While it may be too soon to talk about plug-and-play networking, it is no longer the stuff of science fiction.
Arthur Cole is a freelance journalist with more than 25 years’ experience covering enterprise IT, telecommunications and other high-tech industries.