Open Source Approaches to Traffic Engineering for Cloud

If you’ve spent any time architecting or operating your organization’s network, you’re already intimately familiar with traffic engineering. What might be less familiar, however, is traffic engineering for cloud. As your organization and customers add more cloud resources and hybrid workloads, traffic engineering can get a lot more complicated. Network equipment vendors will tell you they can make that complexity go away—provided you’re willing to lock yourself into their proprietary tools and ecosystems for the foreseeable future. If that doesn’t sound like a great tradeoff (and it shouldn’t), then it’s time to look at open-source alternatives. That’s where companies like Lumina come in. 

As our customers have already demonstrated in real-world networks, it’s not only possible to implement traffic engineering for a wide range of cloud use cases without any proprietary vendor mechanisms, it’s the smarter way to do it. Let’s take a closer look at how you can provision and monitor the network for some of those use cases, using entirely open-source and open-standards based approaches.

 

Traffic Engineering 101

Before diving into the cloud-specific parts of the story, let’s review what traffic engineering is all about and why it is needed in the first place. Fundamentally, operators employ traffic engineering to optimize network efficiencies and provide for deterministic traffic restoration scenarios—to measure and configure them in such a way that they use the underlying resources in the most effective manner possible. They do this by applying technology and scientific principles to the measurement, modeling, characterization, and control of network traffic to achieve specific performance objectives. 

Typically, you’d use traffic engineering to optimize things like packet transfer delay, packet delay variation (jitter), packet loss, and throughput. Issues like these commonly crop up in core networks (or any MPLS network) where packets are routed at each hop, meaning you don’t necessarily know at the source the full path traffic will take to get to the target. For example, because each router makes local decisions, it’s common for variation to get introduced into the traffic flow, where some packets take longer routes than others. For delay-sensitive services like voice and video, this variation can significantly degrade the application experience. By measuring packet flow information and making it available to a system that designs optimal network paths, it becomes possible to reduce these variances. 

That’s just one scenario. Classical traffic engineering gets used for a wide range of purposes, including:

  • Minimizing network congestion
  • Maximizing overall network capacity utilization  
  • Implementing and optimizing fast rerouting in the event of a link or node failure
  • Controlling overall network delay for certain classes of traffic, such as voice or video
  • Identifying traffic from a specific customer for special handling

 

Cloud Adds New Wrinkles

Classic traffic engineering has been an essential tool for operators to control and optimize traffic flows over their networks for many years, and it’s well understood and highly effective for conventional on-premises networks. Today though, as high-speed Internet has become broadly available and the WAN has grown much more reliable, customers are bringing more cloud resources into the mix. That could be public cloud (like AWS or Azure), an operator’s own cloud, a customer’s private cloud, or any other set of compute/storage resources located theoretically anywhere, that can be controlled in a software-defined way. 

When you have visibility into all those cloud and hybrid resources, and you can centrally manage and control them, you can support all sorts of new use cases:

  • Efficiently provisioning and managing distributed Wi-Fi routers and branch office devices
  • Running hybrid workloads, where some services within the workload reside on-premises while others are in the cloud
  • Unifying control of services that span multiple network domains (such as packet and optical networks) 
  • Provisioning services where multiple clouds communicate with each other, such as with inter-datacenter or intra-datacenter services

 

Open Source Traffic Engineering

Lumina’s SD-Core solution provides a comprehensive toolset to implement traffic engineering for these and other cloud scenarios using an open source- and open standards-based approach. You can aggregate control for a wide range of use cases, across multiple network domains, all based on the open source OpenDaylight SDN controller. SD-Core can:

  • Simplify the deployment of new network services like:
    • Data Center Interconnect: Automate real-time provisioning with real-time workflows for dynamic inter-data center use cases
    • Cloud-Connect: Build private network interconnects to public cloud providers such as AWS "Direct Connect" and Azure "Express Route".
  • Optimize the management of existing networks, such as:
    • MPLS/BGP/SR Networks and Services: Accelerate provisioning of backhaul, aggregation, and metro/Carrier Ethernet services like L3VPN, ELINE, ELAN, and ETREE
    • Transport: Simplify management of transmission and optical networks
    • Multi-Domain Networks: Provision end-to-end services spanning IP, optical transport, access, RAN, Whitebox, and MPLS domains.

We simplify the process to capture the key metrics needed to continually optimize services for these use cases, including latency, capacity, packet loss, jitter, throughput, response time, and QoS. You can also draw on the Lumina Extension & Adaptation Platform (LEAP) here to integrate real-time telemetry to validate that the traffic engineering policies you’re implementing are actually having the desired effect. SD-Core supports all the standard protocols needed to implement device configurations (BGP-LS/LU, PCEP, OpenFlow, NETCONF) as well as end-to-end EVPN and VXLAN cases. And, it supports standardized APIs to provision these topologies and manage paths and services.

Traffic Engineering Use Cases for the Cloud

Each of these cloud use cases has specific implications for how you manage and optimize traffic flows. But while the network and use cases may be more complicated, the requirements for traffic engineering are still broadly the same: You need to define service endpoints to originate at a datacenter or on-premises resource. You need to be able to discover the topology of WAN connectivity endpoints. And you need to be able to provision configurations at those endpoints and the end-to-end paths between sources and targets.

 

Open Source-Based Traffic Engineering

Lumina’s SD-Core solution provides a comprehensive toolset to implement traffic engineering for these and other cloud scenarios using an open source and open standards-based approach. You can aggregate control for a wide range of use cases, across multiple network domains, all based on the open-source OpenDaylight SDN controller. SD-Core can:

  • Simplify the deployment of new network services like:
    • Data Center Interconnect: Automate real-time provisioning with real-time workflows for dynamic inter-data center use cases.
    • Cloud-Connect: Build private network interconnects to public cloud providers such as AWS “Direct Connect” and Azure “Express Route.”
  • Optimize the management of existing networks, such as:
    • MPLS/BGP/SR Networks and Services: Accelerate provisioning of backhaul, aggregation, and metro/Carrier Ethernet services like L3VPN, ELINE, ELAN, and ETREE
    • Transport: Simplify management of transmission and optical networks

Multi-Domain Networks: Provision end-to-end services spanning IP, optical transport, access, RAN, Whitebox, and MPLS domains.

Aggregated Control Enabled by OpenDaylight

We simplify the process to capture the key metrics needed to continually optimize services for these use cases, including latency, capacity, packet loss, jitter, throughput, response time, and QoS. You can also draw on the Lumina Extension & Adaptation Platform (LEAP) here to integrate real-time telemetry to validate that the traffic engineering policies you’re implementing are actually having the desired effect. SD-Core supports all the standard protocols needed to implement device configurations (BGP-LS/LU, PCEP, OpenFlow, NETCONF) as well as end-to-end EVPN and VXLAN cases. And, it supports standardized APIs to provision these topologies and manage paths and services.

 

The Lumina Networks Advantage

There’s no escaping the fact that cloud workflows make traffic engineering more complex. But in a world where your customers’ applications and services are likely to be increasingly moving to cloud-based and hybrid infrastructures, the ability to optimize traffic flows is just as essential as it was in legacy on-premises environments, if not more so. Just because it’s important, however, doesn’t mean it’s worth locking yourself into a vendor's proprietary ecosystem to be able to do it. Fortunately, you don’t have to. 

With solutions like SD-Core, you can support practically any traffic engineering use case your customers require, using an entirely open source and open standards-based approach. You can avoid relying on proprietary mechanisms you don’t control, which limit your flexibility to define end-to-end service management. You can employ open source traffic engineering tools, where you have the freedom to customize and update the source code as you choose. And, when your traffic engineering strategy is based on reference protocols and APIs, you retain the freedom to explore any hardware or software option you choose in the future, with any vendor that supports open standards.


Learn more about:

SD-Core

Lumina Network Resource Manager

Lumina SDN Controller

LEAP