Integrating white-box switches and routers into business networks has its challenges, but properly scaled and vetted deployments can save money.
If you’re an enterprise CIO, CFO, or network operations type, you’ve probably been reading about how this service provider or that cloud provider have saved up to 50% on network equipment by using generic “white-box” technology instead of proprietary routers and switches. It’s hard not to wonder whether your own network budget could buy twice as much gear, and what projects might now meet their business case. Could enterprises get in on the white-box revolution? Maybe, if they can address the issues that even service providers and cloud providers have already faced, and in some cases been bitten by.
Compatibility
The first issue is finding the hardware and software. White-box hardware needs software, either an all-inclusive “network operating system” that provides all the features you need, or an operating system plus a separate routing/switching package. The software can’t just be shoveled onto something and run; it has to match the hardware. In some cases, the matching process is facilitated through the same sort of drivers found on PCs and servers, but not all hardware has a driver suitable for all software. Pick a white box and you may not find software you like for it. Pick software you like, and no white box you like may fully support it. This is why self-integration of white-box hardware and software is something enterprises should avoid at all costs.
Vendor push-back, tech problems
The second issue, and likely biggest, issue is introducing white boxes into the network. There aren’t many enterprises worried about network costs that don’t have a current, running, network built on proprietary devices. There are probably fewer who will fork-lift out their current network technology to use white boxes, so white-box switches and routers will have to be incorporated into a network likely built from proprietary devices. That introduces a cultural problem and a technical problem.
The cultural problem is simple; your current network vendor will resent your white-box decision and will likely blame every hiccup or fault you experience on the new gear. When that happens, and you go to your white-box vendor to get their side, they may well blame everything on the software, or they may point the finger back at the proprietary vendor. The software supplier will return the favor by pointing at everyone, and all the players may point at your own integration efforts as the source of the problem. If you had finger-pointing in a two-vendor network, white boxes can make that look like a love feast by comparison.
The technical problem is one of management. All network devices have to be managed, and the management systems and practices most enterprises use tend to be tuned to their current devices. White-box management is usually set by the software, and you can’t necessarily expect much of a choice in how the management features work. That means your current network-operations people will have to contend with multiple management choices depending on the devices they use.
Financial potential
Is there no hope of a white-box windfall for your network and CFO? All of the white-box risks enterprises face can’t be waved away, but there are things you can do to reduce them to the point where benefits outweigh risks. The best place to start is to pick candidate applications for white boxes that won’t get you into trouble.
If you need a switch or router here or there, don’t think about starting a white-box revolution. White-box technology is best used in contained missions, where there’s limited or no mixing with proprietary elements. The most successful white-box projects enterprises have undertaken have focused on the data center, but replacing a bunch of VPN edge routers or installing the LAN switches in a new facility are also good candidates. Keep in mind that there’s little benefit to a few white boxes; you need missions that involve enough devices to make the savings meaningful. Think big enough, but not too big.
Once you’ve found some good, contained missions, you can consider product sources. Enterprises need to focus on getting hardware and software as a tidy package. Celestica, Dell, EdgeCore, Foxconn, Lanner, and Quanta are among the best-known white-box vendors. Enterprises can match the product lines of each with their switch/router needs and pick a vendor who offers not only what’s currently needed but also what might end up having to be added due to changes in traffic levels.
Some white-box vendors may give you a software choice. White box software is available from companies like Arista (the EOS family), Arrcus (ArcOS), Pluribus (Netvisor ONE), and others. All these companies supply a “network operating system” (NOS) that includes all the software you’d need to convert a white box into a switch or router. That includes management interfaces, too, so you’ll be able to establish whether a given solution will fit into your network operations tools and practices.
The final step is trial, trial, trial! Do not rely on representations from suppliers or stories about white-box success. You should plan on a contained lab trial and a field trial before you commit to a full-scale deployment, and get your white-box vendor onboard with that during your initial product assessment. In the trials, you’ll want to test performance and manageability and validate the interfaces between your white-box elements and your existing network. Be sure you have good visibility at those interfaces, because finger-pointing issues will be focused there.
If you still have concerns with integrating white boxes with your proprietary network hardware, check if your switch/router vendor offers “disaggregated” software and hardware, meaning they sell the two separately. That’s not going to save you much money compared to true white-box networking, but it might be worth thinking about if you have a few hold-out proprietary devices complicating a nice new white-box deployment mission. See if you can run the white-box software you selected on those proprietary devices, and you may save a lot of finger-pointing.
If you do things right, what can you expect to save? Enterprises don’t match the 50% savings that some service providers have reported, but those who follow all these rules save about 30% on device costs and have been able to keep management and support costs from rising. So if a 30% savings sounds good, and if you can apply the recommendations I’ve made here, white-box networking is worth checking out. These days, picking a white-box strategy isn’t really any riskier than changing vendors.
