By Casimer DeCusatis
One of the important themes at the 2018 OFC conference was the growing interest in disaggregation of hardware and software. The market for compact modular equipment is among the fastest growing segments of the optical hardware market, particularly among cloud and telecom service providers. But why is disaggregation receiving so much attention, who benefits from this approach, and what lies ahead in this rapidly evolving market?
It is important to recognize just how disruptive this trend has been to global telecom and cloud providers. A few decades ago, the design of optical equipment was dictated by a handful of large companies, and the rest of the industry adopted these form factors to take advantage of low costs that accompany high production volumes. However, the market has changed significantly since then; the hardware and software requirements of Google, Verizon, AT&T, Facebook, and other huge companies are quite different, making it difficult for equipment providers to build components for just one network operations market segment. Instead, there’s a desire to build networks from modular components, which are not coupled to a particular vendor’s software. The ability to run an optical network operating system (NOS) on any vendor’s hardware makes a tremendous difference in many ways. Customers can now select network equipment from different sources and be assured that a shared, portable, open source, modular operating environment would allow this equipment to work together. The ultimate goal is to completely divorce the network operating system from all underlying hardware (line cards, chassis, etc.) as well as to separate applications from the operating system. The major benefit is greater flexibility and faster time to market. In traditional systems, a service provider is relegated to the vendor’s software update cycles, meaning that it can take a long time to introduce new service offerings. Disaggregation provides the agility to enable new offerings that roll out in days or weeks rather than months, and this is critical to establishing new revenue streams for major service providers.
Impact on Major Stakeholders
There are several major stakeholders impacted by disaggregation, starting with the NOS and merchant silicon providers and extending to companies that build software stacks for different applications. With such a large ecosystem, disaggregation will likely happen in stages. For example, in practice open source developers will probably not write low-level hardware drivers. It’s more likely that ASIC designers will develop their own data plane drivers, because the forwarding plane is unique to the ASIC/silicon inside the hardware. This code could then be open sourced as part of a disaggregated ecosystem.
Early steps in this direction include NFV appliances deployed in hypervisors or containers and open ROADMs for optical networks. In addition, just about any control plane application lends itself to disaggregation (any vendor can easily write a YANG model or REST interface with very little effort). A control plane may manage a single or multiple data planes. Those data planes may be co-resident on the same hardware as the control plane, or they may be distributed across a network. The data plane is a bit harder, and packet networks with MPLS or Ethernet are even more complex. Nevertheless, it’s possible to provide a forwarding abstraction layer between the control and management planes and the hardware/software data planes. We can see this trend in recent “white boxes” that use network processor rather than standard CPUs for the data plane. As a result, rather than doing a lot of things reasonably well, these white boxes do only a few things really, really well (namely, network processors are very good at routing high speed packets).
By next year’s OFC conference, expect to see a lot more news about software and hardware disaggregation in the optical infrastructure including the extension of disaggregation to the transport layer.
Does your company plan to leverage disaggregated equipment in the future? Drop me a line on Twitter (@Dr_Casimer) so we can talk about it, and I might use your ideas in a future blog post.
Posted: 27 July 2018 by
| with 0 comments