As you read about in our recent post, Unitas partnered with Equinix and Blue Planet, a division of Ciena, to present the proof of concept: AI-driven Federated Domain Operations for Ultra Resilient Services. Fifteen teams competed, and our team was a winner!
The PoC Showcase at MEF is one of the connectivity industry’s most prestigious awards. MEF itself plays an important role in the industry: it is the global Ethernet standards body to which service providers worldwide adhere, and it’s what allows providers and their vendors to interconnect and provide end-to-end service—since no one provider has a network everywhere. The Unitas, Equinix, and Blue Planet team competed against players globally and advanced to win the Orchestrated Network Service Implementation award.
The Inspiration: Why the Market Needs AI-driven Federated Domain Operations
With the increased usage of video streaming, gaming applications, and data intelligence, the demand for ultra-resilient services with 100% network availability with granular reporting of service performance grows in turn. These applications often rely on cloud and network services running over global interconnection fabric architectures with distributed virtual and physical network functions in multiple operator domains. AI-driven network management replacing reactive fault-driven network management is essential for these new ecosystems.
AI-driven network management will enable the prediction of SLA exceptions, re-routing, optimization, and other remediation actions on a very large scale. This in turn will drive increased use of cloud-based services for mission-critical applications and increased innovation.
Without an intelligent global fabric of network and cloud access—including last mile through the cloud edge data center—there will be gaps in service and performance. Providers will struggle to sustain 100% service availability of applications and data, and to support new services and applications that require multiple domains, last mile access, and edges with VNF and cache mobility.
To address this increasing demand for global and ultra-resilient services, Unitas collaborated with Equinix and Blue Planet, a division of Ciena, to build a global network fabric with proactive AI-powered intelligent automation that can support such services.
Downtime with Increasing Network Complexity
To set the scene, networks today are incredibly and increasingly complex in how they are configured. The slide below illustrates the end-to-end view of a network, starting from the end client premise through carriers to the Unitas Cloud On-Ramp, through Equinix’s Global Cloud Exchange Fabric (ECX) to the edge.
The present problematic modes of operation are:
- Disjointed: domain silos have poor handoffs and poor or no visibility across providers
- Slow: it is a multi-step, manual, and complex triage to resolve issues with service, leading to high OpEx costs
- Finger pointing between operational teams from domain to domain (or network to network), resulting ultimately in customer dissatisfaction, slow service, and slowed adoption rates
Our team’s PoC focused on a future where networks have:
- Visibility of service performance across ecosystem, leading to deeper insights for proactive customer care
- Speed to troubleshoot and isolate service issues means increased adoption
- Proactivity with AI-driven operations across service ecosystem, resulting in reduced service calls and reduced OpEx costs
What It Means to Leverage Automated Intelligence (AI)
Today, tremendous amounts of log data and information are being gendered by network elements—networks which are comprised of many different domains. Reacting to a red-light warning of a problem within your network means you are already have an issue, since you are working to restore service. It would be ideal to work proactively, seeing trends of congestion and utilization in the network before an impacting event occurs.
Our team’s PoC used intelligence to address congestion challenges before it impacts the user experience or there is an outage. From a machine learning standpoint, the AI technology we use correlates log data and identifies warnings across millions of data points. It computes millions of network data points in order to understand particular scenarios that causes congestion, the system proactively sees the potential for an outage to occur.
One thing to remember is AI does not happen in a vacuum, ideal results are not machine intelligence and action alone: human interaction plays a key role. AI correlates, filters, and synthesizes network information to propose actions to the human user. The user either approves or denies each action, which creates learning events for the machine. For actions that were denied, the machine “learns” not to propose those actions again. Over time, the AI learns to provide better and better actions, and the network performance continues to improve.
These AI capabilities are helpful because they far exceed what a human can see and react to yet are honed by human oversight. The more data points, logs, and information we have on the networks, the more we can work with AI to correlate and analyze to ensure network uptime. Blue Planet provides the AI capability to our joint solution. Using advanced analytics and AI to sort through these warning points is critical for any level of support for the modern enterprise, particularly with multiple networks and technologies involved.
This diagram (below) shows an example of a multi-network (federated domain) and its flow across our and our partner services.
A Unitas business customer is using digital services from a content cache at an Equinix data center. The congestion points along this route can be highly variable. Content is available from core servers as well as regional edge caches. Blue Planet’s AI monitors and makes recommendations to the operations engineer in the CMC (Cloud Management Center the NOC at Unitas Global) to proactively reroute traffic from a core cache to local cache before congestion is experienced. AI makes recommendations across the multiple network domains to maintain the quality of the content and availability of the network.
The Significance of This PoC
The PoC that Unitas, Equinix, and Blue Planet presented is important as it demonstrates that we could create a global interconnected network that is resilient against failure or congestion from end-to-end. Applications work seamlessly and are impervious to disruption with this kind of orchestrated multi-network service.
Today, Unitas offers an interconnected multi-operator, software defined fabric delivering a robust underlay to support demanding cloud applications and workflows. As the adoption of interconnected network fabrics and hybrid cloud services accelerate, enterprises and managed service providers face an urgent need to use intelligent proactive methods of inventory management, service activation and service assurance across untrusted and cooperating domains. Without an intelligent global fabric of network and cloud access, including last mile through the data center edge, there will be gaps in service and performance and providers will continue to struggle to sustain service availability of applications and data and to support new services.
Thank you to all the team members from Equinix, Blue Planet and Unitas Global. We’d like to extend a special thanks to Tony Thakur, our SVP of Connectivity, who lead Unitas’ participation in the team. We would also like to thank Chris Roeckel and Jose Ladeira from Unitas for their contribution to the MEF PoC.