High-Performance Networking from Edge to Core


By: R. Scott Raynovich

This Futuriom Leadership Brief is sponsored by Graphiant. You can download it here.

Recent Futuriom MCN and hybrid cloud networking research has identified some new needs. Feedback from the marketplace indicates that most end users likely won’t use traditional networking tools to connect to multiple or hybrid clouds. Why is that? Quite simply, traditional networking gear, such as routers and switches, was built for single datacenter or private enterprise environments.

The multicloud or hybrid network must adapt to the ephemeral nature of the workforce and applications. It can’t be tied to specific boxes, networks, or even IP addresses – and it must be quickly programmable. It must also provide built-in, zero-trust security.

Adapting Networks to the Edge and the Cloud

Let’s look at several of the key limitations of traditional networking gear for hybrid cloud or public cloud applications.

  • IP Transit limitations. Organizations increasingly rely on IP Transit and the Internet to create overlay tunnels that bridge the gaps between cloud services, but there is no common or quick way to define and implement service levels or policy across multiple clouds and IP Transit networks.
  • MPLS is past its time. Network innovations such as MPLS were effective for connecting private datacenters, but they were built for a different architecture based on branch offices, headquarters, and datacenters – not the ephemeral hybrid workforce, which might be roaming across different locations and consuming apps from all corners of the world. MPLS services are also very expensive, and IP transit is becoming an increasingly affordable alternative, if implemented correctly.
  • Adoption of infrastructure as code and APIs. Network managers want to move toward “infrastructure as code,” which enables them to build networking policy directly into applications. This requires tools that can talk to public cloud networking constructs and be programmed via application programming interfaces (APIs) at the application layer.
  • Scripting has limitations. Past enterprise datacenter implementations are often scaled and managed with scripting tools, such as Ansible or Terraform. However, these scripting tools are largely operated by humans in response to scaling demand. The answer is to build an automated network that can respond in real time to demand for high-performance IP Transit.
  • IPsec can only do so much. The generic IPsec tunnel overlay has become the standard for connecting across cloud networks. However, each of these tunnels must be provisioned individually. “Tunnel scaling” can become a problem in many cloud-scale networks. In addition, IPsec tunnels may consume excess resources by requiring a lot of encryption and decryption.
  • Networking impact on cloud costs. Cloud egress data costs present another challenge to building networks using cloud resources.

The bulk of end users surveyed by Futuriom indicate they envision using a hybrid network that can connect multiple environments or hybrid clouds. In many cases, networking for hybrid and multicloud arrangements will prove to be a major challenge. Connecting different computing domains or platforms requires a flexible networking infrastructure that can also be managed and monitored to ensure security and compliance.

Enter the High-Performance Network

Hybrid networking needs to be flexible enough to connect to many different types of networks and devices, and it needs to be programmable. But there are many other needs. If one were to sum up a key characteristic of what organizations need from networking, it’s the same thing they got when they moved to the cloud: a more responsive infrastructure that could respond on demand to business needs and new service models.

These new business service models require new cloud networks to execute on the following:

  • Connection of any network or device: This means the capability to connect to any network, any application, or any device, including edge compute, Internet of Things (IoT), cloud, and enterprise networks.
  • Full application-level programmability: The capability to program and configure using automation driven by data models and APIs in order to respond in real time to business needs and applications across a multicloud or hybrid cloud network.
  • Unified networking visibility: The capability to provide a simplified management experience and single networking fabric for connecting to any network topology or domain, ranging from enterprise to hybrid and multicloud.
  • Cost-effective hybrid cloud networking: Providing a cost-effective solution for transporting data among clouds, including consideration of the economics of costs such as cloud egress charges.
  • Full policy and security control: Centralized network security and policy control to provide quality of service (QoS), secure end-to-end encryption, and threat mitigation.

Graphiant’s Solution: A Stateless Core Cloud NaaS

Graphiant has introduced in interesting solution to these hybrid networking challenges: The Graphiant Network Edge is a solution that can solve these challenges while putting policy, control, and QoS fully into the hands of any organizations looking to operate an MCN.

Futuriom has written a Leaderhip Brief on this exciting new technology. Go ahead and download the Leadership Brief now.