As 2013 comes to a close, it’s time to pull out the crystal ball and make some predictions about 2014—specifically, about what we’re likely to see in the Network Virtualization space in the coming year.
The safest prediction I can make is that Network Virtualization will see a significant uptick in adoption next year. That’s based in part on a simple extrapolation of what we’ve seen in 2013. We’ll end 2013 with significant customer momentum for the VMware NSX network virtualization platform. Adoption has started to spread beyond public cloud providers and technology-driven customers like eBay to large enterprises, notably the large financials. So it’s a safe bet that 2014 will be a year in which adoption moves even further into the mainstream. We’ll also see many of these customers moving from proof-of-concept trials to live production deployments.
Expanding on the spread of network virtualization into other markets, we’re surely going to hear a lot about “Network Functions Virtualization” (NFV) in 2014. A full discussion of NFV (hint: it’s not the same as network virtualization) is going to have to wait for another blog post, but it’s clear that the telcos have put a lot of thrust behind this initiative. NFV promises to virtualize many of the functions that have historically been implemented as vertically integrated systems sold by traditional network equipment providers. For example, many of the components of an “evolved packet core”—the heart of a 4G wireless network—are amenable to implementation as virtual machines on x86 rather than as vertically integrated, single-purpose boxes. And just as the need for network virtualization became apparent once server virtualization took off, NFV is also going to drive requirements for network virtualization. Once those functions are implemented in VMs, there will be a need for a way to “chain” services together and to dynamically create topologies that interconnect virtualized functions. So I’m predicting not only that NFV will be getting a lot of airtime in 2014, but that the importance of network virtualization in the NFV context will become much more apparent.
As we look at the increasing adoption rate of network virtualization, we’ll also see a wider range of reasons for adoption. Today, the number one driver for adopting network virtualization is agility. That will continue to be true for a while, but one effect of greater adoption of the technology is that customers will find new motivations beyond the most obvious. The same thing has already happened to server virtualization: the initial use case was workload consolidation, but over time other benefits of virtualization (such as VM migration, ease of provisioning, etc.) drove adoption. We don’t know all the use cases of network virtualization yet, but we’re going to learn more in the coming year.
One example of what we’ve seen in 2013 will give a sense of what’s possible. A customer wanted to deploy a new NSX feature – distributed virtual routing – but was unsure how it would impact his production environment. So rather than testing it in the customer network, we had the customer take a snapshot of his network and send it to us. We then applied the configuration change to enable the new feature, tested it in our lab, verified that everything worked as expected, and sent a snapshot of the new, reconfigured network back to the customer. This illustrates the transformational power of network virtualization: the entire state of a network is just a set of bits, which, just like a VM image, can be snapshot, cloned, shipped around, etc. So I’m predicting that this sort of use case will become much more common as network virtualization matures in 2014.
Another trend that is sure to continue in 2014 is an increased focus on higher layers of the protocol stack. It is fundamental to network virtualization that the whole stack, from L2 to L7, needs to be virtualized, so that virtual networks faithfully reproduce the capabilities of physical networks. While a great deal of work has gone into delivering robust capabilities at L2 and L3, those lower layers are becoming “table stakes”. What will really differentiate various network virtualization solutions in 2014 will be the support of higher layer services such as load-balancing, firewalling, and security services.
Furthermore, higher layer services will be another area where we discover the real potential of network virtualization, as these higher layer services leverage information that’s available in the hypervisor. For example, the hypervisor can detect the applications running in VMs, and the users logged in to them, and pass that information to security services, allowing much richer capabilities than could be implemented in a traditional network-based security appliance. We’ll also increasingly see these services driven directly by higher-level policies—network configurations will be expressed in terms of a business objective rather than in terms of low-level details like IP addresses and port numbers. (In a related development, we’ll spend less time debating tunneling protocols—I hope!)
Finally, we made a good start to establishing a network virtualization ecosystem in 2013, and this will solidify and expand in 2014. We have a collection of hardware partners whose devices enable NSX to manage physical ports in much the same manner as virtual ports—their products will become generally available in 2014, and the set of features they support will become richer. We’ll partner with providers of virtualized network functions in support of our NFV activities. And the all-important L4–L7 services mentioned above will be greatly enhanced by deeper integrations with partners who deliver those services in both virtual and physical form factors.
All in all, there’s a lot to look forward to. I don’t think I’ve made too many risky predictions here, but the best bet is that network virtualization will enable some new capability that I haven’t predicted, because we are clearly still discovering all the potential of this technology.