Three Key Insights for Building and Maintaining Disaggregated Clusters

Disaggregation and DriveNets

DriveNets leverages a Distributed Disaggregated Chassis (DDC) architecture to build any router type, from core to edge, by clustering dozens of white boxes that function together as a single routing entity. One of the core elements of this architecture is disaggregation, which is the separation of hardware and software components that carry out the main functions of a network on switches, routers. Disaggregation eliminates vendor lock and offers SPs a hardware-agnostic network solution.

Building our first DDC cluster was challenging. However, with experience garnered from hundreds of large-scale deployments across Europe, North America, India and Japan, we’ve refined our operational processes and gained valuable insights. Today, new deployments are significantly faster and require minimal integration efforts.

While a full DriveNets Network Cloud training overview is beyond the scope of this blog post, I’d like to highlight three key insights that simplify the process of building and maintaining DriveNets DDC clusters, ultimately easing the path to building networks like cloud.

#1 Planning is everything

DriveNets Network Cloud’s distributed design offers power, space, and heat advantages over centralized chassis-based architectures. However, effective planning is still crucial to fully leverage these benefits. By following simple steps, we can mitigate potential limitations and maximize the advantages of a disaggregated and distributed cluster.

Site survey

During site surveys, we must assess power configurations and available rack space. This simple step helps us plan a more efficient cluster design. DriveNets’ distributed cluster design allows us to utilize existing rack space, minimizing the need for additional racks and reducing cable costs and power consumption by using multiple cable types (AOC, AEC, or DAC).

Rack distribution

We should keep growth in mind and design rack distribution in a butterfly architecture so that the fabric elements (Network Cloud Fabrics, or NCFs) are placed in the middle. In this way they can reach optimally any current and possible future network elements (Network Cloud Packet Forwarders, or NCPs) regardless of their location. As a result, we can leverage the cluster’s practically unlimited scale capability and still utilize rack spaces and cables more efficiently.

Deploying a disaggregated cluster may seem as simple as assembling cables and white boxes. However, proper planning and execution are crucial for optimal performance and scalability.

#2 Plug-and-play during normal operations is realistic

After a cluster is built and operational, there may arise an occasional need to replace or add a new NCP or NCF. DriveNets Network Cloud’s inherent redundancy and carrier-grade resiliency ensure the cluster’s continued operation, even with a failed NCP or NCF. This means that replacing or adding a new NCP or NCF is done by simply plugging out a malfunctioning white box, plugging in a new one, and allowing it to automatically configure and go online.

Should this plug-and-play capability be implemented during normal operations or a maintenance window? While the plug-and-play capability works flawlessly, it’s essential to consider the specific cluster’s capacity and redundancy design. If, for example, the capacity is stretched, replacing a device during a maintenance window can prevent potential blackholing.

Tier-1 service providers often restrict IT work to maintenance windows, limiting flexibility. However, adding or replacing a plug-and-play white box during normal operations offers significant operational advantages, which should be leveraged whenever possible.

#3 DNOR’s impact – from skepticism to excitement

With each new operations training, we hear initial concerns about deploying large, disaggregated clusters. Accustomed to monolithic solutions, operators often find the idea of deploying dozens of white boxes from one vendor and software from another to be daunting. However, DriveNets Network Orchestrator (DNOR) quickly changes their perspective.

DNOR simplifies this process with autoboot profiles, streamlining the provisioning of new large clusters and enabling zero-touch deployment. This allows service providers to enjoy the operational simplicity of monolithic solutions while benefiting from the advantages of disaggregated network routers.

Additionally, DNOR’s smart rollout feature automates software upgrades and patches across the entire network, adhering to pre-defined timeframes. This eliminates the need for manual site-by-site software updates.

The more SPs use DNOR, the faster we’re seeing a shift from skepticism to excitement. SPs increasingly value DNOR’s ability to streamline operations and empower their teams to achieve more with fewer resources.

Disaggregation with frictionless integration

SPs understand that disaggregation is a necessity that cannot be avoided. Ultimately, the current monolithic approach is no longer viable, and a move towards disaggregation is essential. We’re proud to be part of the major network changes that large SPs, such as AT&T, Comcast, KDDI, Turkcell, and others, are making across their network domains. The global success of our DriveNets Network Cloud solution enables us to offer SPs the advantages of disaggregated networking while eliminating “scary” deployment challenges.