Carriers Driving the Need for Disaggregated Networks
At Mobile World Congress (MWC) 2016, John Donovan, then the Chief Strategy Officer and Group President of AT&T Technology and Operations, stated in a fireside chat that white box networking would be the single biggest change in the telecom industry with the most profound impact on how we design, build, and operate networks.
Interestingly, the poll conducted in the public attending the discussion gave zero votes for white box networking.
In 2022, AT&T announced that it had reached a critical milestone of 52% of its network being virtualized, using standard white box hardware running software that reproduces most networking functions, and that the carrier is targeting 65% of its traffic running on disaggregated architecture by the end of 2023. From core to the edge, AT&T’s networks, which deliver services to almost 200 million customers, are running on white boxes.
Why has AT&T been so relentlessly driven all these years to transform its networks into a Software Defined Networking platform?
As John Donovan bluntly put it: “(It’s a matter of) survival.”
Customer expectations and the advent of “non-traditional” Over the Top (OTT) competitors created a mega pressure on communication service providers (CSP) to deliver innovative services quickly and in a profitable manner to ensure a growing bottom line and continuous investments in the network.
The genius of John Donovan and AT&T as a whole was to recognize the source of future innovation — not the old traditional networking model, where a handful of vendors such as Cisco and Juniper would lock you in with closed and proprietary systems that are unable to communicate with the flourishing and ever-growing ecosystem of enterprise and consumer applications. The “old” way does not work any longer. Although sufficient to push packets from one router to another, traditional networks fall short of new requirements. The network has to be in sync with the upper layers, understand in real-time the requirements of user applications, and allocate the necessary network resources to ensure an optimal quality of experience.
At the heart of white box networking (also known as open networking or disaggregated networking) lies the network operating system, the cornerstone of innovation. A disaggregated network operating system enables any third-party application to communicate with the network and send critical information to ensure the network delivers the appropriate bandwidth, roundtrip, and jitter times for an optimal and constant end-user quality of service (QoS).
The role of the open network operating system is to coordinate all routing decisions taking place on the control plane and execute these decisions on the data plane to forward traffic using the correct routes. But there is a newer role of the network operating system (NOS), which is almost non-existent in traditional networking. It is about changing the network into a “business-aware” entity, understanding the context in which traffic is forwarded from point A to point B and applying the rules that will guarantee the best conditions for all applications.
These decisions must be made at scale because tens of terabits per seconds of traffic are continuously flowing from the peering points all the way down to the access networks through the core.
Requirements of a Disaggregated Network Operating System (NOS)
To meet modern networking needs, there are several requirements of the NOS:
- It has to provide a wide array of protocols and features
- Its core software architecture must mimic cloud platform architecture
- It needs to be open
- It has to be fully secure
ExaNOS, Exaware’s advanced open network operating system matches all of these requirements.
Network Operating System for Disaggregated IP/MPLS Router Solutions
Support for a Wide Array of Protocols and Features
CSP networks are incredibly complex in nature. They run a large set of protocols, above which many features are applied, to ensure that network traffic is handled in the right way and fully secure.
Internal networks run an Interior Gateway Protocol, usually OSPF or IS-IS. The exterior Gateway Protocol is traditionally BGP. You might also need to run multicast in some parts of the network, especially if you stream video. In addition, you need to deal with IPv4 and IPv6 addresses. You need to filter traffic on your peering points, which requires scalable Access Control Lists or even BGP-Flowspec if you want to automate DDoS attack filtering.
To meet the needs of CSPs, the requirements of an open NOS are complex.
ExaNOS supports all protocols and features needed by CSPs, enabling you to run the same NOS across your entire network.
A Core Software Architecture that Mimics Cloud Platform Architecture
Cloud elasticity, the ability to scale up and down cloud resources, is also a key performance factor in open networking.
Hardware resources need to be available, but the ability to coordinate all protocols and features as well as apply instant routing decisions and quality of service priority queuing at scale requires a flexible and robust NOS. Cloud computing has demonstrated its ability to scale almost infinitely. An open NOS needs to replicate the same architecture: distributed internal processes which focus on a few tasks and work concurrently to deliver the traffic scale required at any point in time.
ExaNOS, which ran in Tier-1 production networks at Comcast and NTT embeds in its core a distributed architecture, which can be replicated on every single line card of large hardware routers.
The internal “fault protection” mechanism handles any exception to isolate it from other processes and ensure a nonstop routing traffic.
ExaNOS can scale up to hundreds of terabits per seconds, running the full gamut of protocols and features of network operating systems.
Open Network Operating Systems
The number one criticism of the old networking model of Cisco and Juniper is its closed architecture. Every time you buy a new router, you need to buy and pay for the NOS again — you cannot reuse the licenses you bought in the past with different hardware. Traditional network operating systems include tons of features you do not necessarily need but are there for legacy purposes. However, the interesting features such as L2/L3 VPN incur an extra charge, based on the bandwidth required.
The reason why IP networks have not innovated since the nineties is because traditional vendors have purposefully kept their “walled gardens” completely closed. You get, at best, the parameters available in the SNMP MIB, more aimed at monitoring purposes.
When it comes to deeply embedded third-party applications in the network, accessing the core functions and parameters of the NOS is key. You do that by providing a northbound interface to the operating system, so that your applications can both read routing parameters and modify them at will.
Consider DDoS mitigation as an example.
DDoS mitigation is a fundamental network security service, which could take a massive advantage from the information contained in the peering routers and could also use their filtering capabilities to keep malicious traffic outside of the network.
Although there are existing solutions using BGP Flowspec, their capacity to scale to thousands of simultaneous filters is somewhat limited because the Flowspec interface has its limits. Also the routers themselves can only handle a few hundred filters at best.
With an open NOS, DDoS mitigation capability can be embedded within the whitebox router (but also kept outside of it) and become a true shield against outside attacks, with scale, accuracy, and in real-time. When deployed at every peering point, the DDoS protection systems get to a new level, leaving the internal networks clean of “polluting” traffic. The DDoS identification component can directly communicate with the router’s NOS and instruct it to block, throttle, or rate limit specific source IP addresses.
All of the key attributes of an open NOS should not come at the expense of reliability and security. The open nature of the disaggregated network operating system can raise questions on the integrity of the router and, by extension, of the entire network.
In an open NOS such as ExaNOS, the basic, yet fundamental features of IP routers are still present, such as AAA/TACACS+ and SSH authentication and connection methods.
Security functions, such as MD5 authentication for routing protocols is also available on ExaNOS to ensure that established connections with external routers cannot be compromised.
Control plane ACLs protect the router from unnecessary traffic and ensure the correct prioritization of routing decisions to keep an optimal performance.
Add New Services with an Open Network Operating System
A new chapter of the networking industry history has begun. What many CSPs dreamt about during past decades has now come to fruition. The fundamental building blocks of open innovation in the network are here. Expect to see efficiency improvements in operations using applications such as BGP routing optimization or telemetry to identify network issues before they become customer impacting.
Revenue generating services have become a reality, including on-demand bandwidth and VPN services, personal DDoS mitigation protection per business, and ultra-HD video services where the network guarantees the Quality of Experience (QoE).
To learn how you can leverage disaggregated network operating systems to reduce your costs up to 50% and add revenue-generating services, contact an open networking specialist today.