by Rob Sherwood
If there’s only one message you take from my ONUG white box tutorial, it’s this: reducing capital expenditures (CapEx) is not the only reason to move to white box, branded white box (“brite box”) or disaggregated open networking switches and routers. All of these terms translate into one thing: switches that enable you to buy the hardware separately from the software.
This year is my third time giving ONUG’s white box tutorial, and the exclusive focus on CapEx remains the biggest point of confusion. In my tutorial, and summarized in this blog, I’ll show that while the CapEx savings are in fact huge (when using proper apples-to-apples pricing, see below), there is less often discussed but still real, longer-term value in white boxes from:
- – Reduced operational expenditures (OpEx)
- – Additional revenue from increased business agility
- – De-risking your hardware and software investment
So whether you come for the CapEx or stay for the OpEx, business agility, and reduced risk, white box switching is a technology that you should be investigating.
In a data center deployment, OpEx costs (e.g., time spent monitoring, debugging, and updating network configuration) typically dominate CapEx spending manyfold. White box-based networks help make your existing staff more efficient because you can deploy your choice of modern white box Network Operating System or “NOS,” such as any of Big Switch Network’s SDN OS Big Cloud Fabric, Cumulus Network’s Cumulus Linux, or Pica8’s PicoOS. Rather than the one that’s hard-wired with the proprietary hardware, you can pick the modern NOS that best fits your requirements, and thus optimizes your operational needs. For example, many modern NOSs provide network automation using either an integrated management controller (with Big Switch Networks) or if you have the skill and time, by writing your own (using Cumulus or Pica8).
Depending on the NOS, automation allows you to monitor your networking from a single point, programmatically debug your network, and avoid downtime from configuration inconsistencies — all improving your staff’s efficiency. Additionally, different from a traditional NOS, modern operating systems don’t include 20 years or 20+ million lines of legacy code (RIP, TRILL implementations, DECNET support, etc.) that further complicate deployments.
In talking to customers, it is clear that a major pain point with their networks is that they cannot change them fast enough to keep up with their new applications and workloads; in other words, their networks need more agility. A typical user tells me that deploying a customer facing (read: revenue generating) three-tiered web-application typically can take three to six months to update their network. That is the time needed after the development team has written, debugged, QA’d (quality assurance), and certified the application; it still takes a few months for the network team to file the right tickets, wait for the right maintenance windows to configure the required switches, routers, firewalls, and load-balancers before the company can start realizing revenue from the new application. So, in addition to making your network cheaper to operate, features of a modern white box NOS, such as automation and programmable L2/L3 boundaries, can actually accelerate the rate you deploy new applications, and thus increase total revenue.
Third, white box switches de-risk your investment by removing vendor lock-in between your network’s hardware and software. Many customers are still running end-of-lifed equipment for which they would gladly pay a support contract but cannot because their tightly-coupled hardware/software vendor is disinclined. Another risk is if the traditional vendor’s strategy changes, either through new technology or start up acquisition, and starts promoting a new product line, then your existing hardware stops receiving the latest, best software features.
By moving to white box switches, where you can install the software of your choice, you now have multiple software vendors as options. If one software vendor stops supporting your hardware, at least you have options to move to another. Similarly, if you like your software but the hardware needs replacement, it’s possible to support the same software on a larger variety of new hardware. Also, similar to the server world, hardware/software disaggregation in networking drives higher feature velocity because the components can evolve independently. In other words, white box de-risks your hardware and software investments by removing the single point of failure of the vertically integrated hardware/software vendor.
Last, even though CapEx savings is a relatively smaller part of the IT budget, it can still be quite significant. While there have been a few public TCO surveys of white box versus traditional switching CapEx costs, many of the ones I’ve seen compare a mix of rare, large deal pricing of traditional switches and list pricing of white box switches — to the advantage of traditional vendors. A true apples-to-apples comparison would be list price to list price as all vendors discount, and the level of discount depends on the importance of the deal and not the technology. Our own comparison has shown white box TCO reductions in the range of 40-60% over the traditional switching hardware.
While these thoughts are my own, notable data center architects including James Hamilton  from Amazon and Albert Greenberg from Microsoft’s Azure  have been saying this for years. However, somewhere between what’s easy to quantify (CapEx), apples-to-oranges TCO math, and naysayers’ attempts to dismiss white boxes, the in-depth value behind white box networking has been getting lost in the noise. So remember, while the CapEx savings are substantial when comparing apples-to-apples, there is also additional, longer-term benefit in OpEx savings, increased revenue from improved business agility, and de-risking your hardware investment by removing vendor lock-in. In other words, white box switches are a win for all environments and for many reasons.
The White Box Networking tutorial will take place from 9:30am-12:30pm on May 12th as a part of ONUG Academy at Columbia University.
 James Hamilton, et al: “The Network is In My Way”: http://perspectives.mvdirona.com/2010/10/datacenter-networks-are-in-my-way/
 Albert Greenberg, et al: “The Cost of A Cloud”:
Rob serves as the CTO for Big Switch Networks, where he spends his time internally leading software architecture and externally evangelizing SDN to customers and partners. Rob is an active contributor to open source projects such as Switch Light and Floodlight as well as the Open Compute Project. He was the former Chair of the ONF’s Architecture & Framework Working Group as well as vice-chair of the ONF’s Testing & Interoperability Working Group. Rob prototyped the first OpenFlow-based network hypervisor, the “FlowVisor”, allowing production and experimental traffic to safely co-exist on the same physical network and is involved in various standards efforts and partner and customer engagements. Rob holds a PhD in Computer Science from the University of Maryland, College Park.