The Network is Ripe For Disruption
It’s been a helluva ride in enterprise communications. We blew up the PBX, switched from circuits to packets, embraced smartphones, virtualized the whole thing, and reinvented outsourcing with the cloud.
Everything has changed except the underlying network.
That’s not quite true – we did downgrade the network. TCP/IP, invented in the 50s, is the solution to every twentieth century networking problem. Many forgot how technologies used to come with their own network – SDLC, DECnet, Netbios, NCP, and of course a whole collection of TDM interfaces for telecommunications.
All of those solutions were optimized for specific applications. Fit for purpose, yet we threw it all out. TCP/IP was powering the emerging Internet, and became the lingua franca of networking technology. The world was so busy reaping the benefits of convergence that it ignored the shift away from fit for purpose.
For example, voice over IP (VoIP) enabled things like unified messaging. Another early Greatest Hit was click-to-dial. These were possible because our voice systems could now speak data. Revolutionary benefits with minor inconveniences such as having to force synchronous traffic down asynchronous pipes. Voice professionals had to learn new tricks such as troubleshooting jitter, latency, and one-way audio.
Convergence solved slightly more problems than it created. Unfortunately, the new world was built on an old foundation. TCP/IP is old, and routing has not changed in 25 years. The problem gets worse as our networking requirements evolve. The sheer size of the mess we created makes it extremely difficult to do undo.
Modern networks are encumbered by layers of band-aids. IP routers did not provide sufficient security so we complement them with firewalls. Unfortunately, firewalls were not designed for real-time communications so we added SBCs. Load balancers were introduced because IP is not good at distributing traffic. Failover solutions introduced the need for SDN in the data center. Even VoIP, which was the solution, needs to be repackaged into MPLS frames.
The network is ripe for disruption. It’s the only piece of the IT solution that hasn’t seen significant price/performance gains. Not only has the pricing model remained intact, but equipment prices have actually increased. The cost of compute, storage, and bandwidth have declined sharply.
There are several indicators that the disruption has begun. Here are seven shifts to watch:
SD-WAN: Software defined networks have significantly impacted the data center. They are spreading across the enterprise now, but the next big act will be SDN-WANs. This is especially true for real-time services such as UCaaS.
Open Networking: Traditional networking solutions have their benefits, but they come with bottlenecks, wasted capacity, and premium prices. Many large cloud providers are moving away from name-brand equipment and creating their own solutions – some of which are available as open source.
NFV: I never really understood the difference between a CSU and DSU, and I’m not sure anyone else did either. The revised model is to just put a bunch of boxes (er… servers) and give them whatever role that happens to be needed at the time. Network professionals have to keep changing hats, why not let the gear do the same. NFV has huge benefits for data centers.
IPv6: We have officially run out of IP addresses. The impact isn’t widely felt because our devices share the single address of the router. But think about the number of routers in your life. There used to be one at the office. Now there’s one at work and at home. There’s also one on your smartphone (if you tether). There’s likely one now or coming to your car, too. That’s just linear growth stuff. The real growth comes from IoT. What if we want to give an address to everything? The Internet is about to explode with trillions more devices, sensors, and actuators very soon. IPv6 will support 340 trillion trillion trillion addresses which seemed ridiculous to me a few years ago. Today, it seems right. We will outgrow it, but IPv6 will last at least as long as the current IPv4. Does your equipment (including IP phones) support IPv6? Probably not.
Network as a Service: The network remains the sole component still under full control of IT - and it shows. The world has shifted to opex, agile, and scalable, making our current private networks a bit conspicuous. The network consumes 12-15% of the enterprise IT budget, but users and servers are moving offsite. Networking engineers are in high demand, and training is expensive. That sucking sound you hear is the cloud – it’s already got compute, storage and business applications. Architectural shifts have begun to move management and control to the edge. It seems inevitable that networking itself will become a service.
Trust: Trust no one. That’s probably from the X-Files. The correct term is Zero Trust, and it’s a paradigm shift. The current distributed networking model requires trust among “friendly” devices that propagate important updates. It’s a nice model, but any kid who’s played the game of telephone can tell you there’s inherent risk to such design. The network is only as strong as its weakest link and there’s a lot of links. Zero trust security is the new model. Trust was so last year.
Wireless: It used to be so simple – 802.11 was wireless. Not so simple any more. I was planning to upgrade 802.11n, but it seems 802.11ac is the new orange. Cisco is now using 802.11r to create a “fast lane” for iOS10 devices. Wireless is no longer the fries to the wired network’s burger. Wireless is rapidly becoming the whole enchilada of corporate networks. Many PCs don’t even come with wired ports any more. It’s not just LAN, either – expect big changes with LTE and VoLTE. How many femtocells are in your 2017 budget?
Yes, the network is overdue for big changes. Above were the emerging technologies, below are companies to watch:
- 128 Technology
And of course, Cisco, which has the difficult task of managing innovation along with the innovator’s dilemma.