Analytics and Closed-Loop Automation for Ultra Low Latency Networks
A leading financial services provider that delivers real-time information to global banks needed to introduce network analytics and closed-loop automation that delivers capacity management, performance and latency optimization, predictive analytics, network assurance, and SLA compliance.
Anuta networks, we’re based in San Francisco Bay Area with customers and partners worldwide. Some of our leading customers as you can see here, very large Telcos and enterprise customers all over the world. And what we offer is called closed-loop automation that delivers orchestration, analytics, and telemetry for multi-vendor networks. We support 45 different vendors, and we orchestrate and automate networks across multiple domains whether it’s your data center, Campus, branch, the MPLS core network, and the SD-WAN as well.
Customer Profile & Requirements
So the case Sturdy today is about a financial services customer. They deliver financial news – stock data, as well as disclosure data to their customers and their main challenges, is about delivering this financial data within milliseconds to their customers. And their network engineers said “our network is a slave of the applications. Whatever the application wants, the network has to respond within milliseconds.” So they want to implement real-time network analytics, and they want to make the network very responsive to the application demands, and they want to do this within 500 milliseconds. So we’ll go into details as to how the Anuta’s ATOM is helping him deliver this demand from the applications.
So – a little bit more in detail; the requirements. The solution, the automation should be able to collect the data whether its streaming telemetry or traditional SNMP based data. And this all has to be analyzed and has to create reports and notifications within milliseconds. And this has to scale to thousands of devices and millions of interfaces and this has to be stored for more than a year for compliance purposes. And they want to make sure the platform is fully redundant, and there is no single point of failure. And this has to support open standards.
Analytics and Closed-Loop Automation: Interface Flapping
So the solution we offer as I said is closed loop automation. It’s best to explain using an example, let’s say you can define the baseline behavior for any network. Let’s say you’re looking at the interfaces – you can say I don’t expect this interface to flap more than two times in a day. Now you can define what you mean by that metric and how you want to collect it and how frequently you want us to monitor it.
And then the correlation engine is comparing the current behavior with the baseline behavior and based on the various deviations that are happening; it can automate the corrective actions. So for example, if the interfaces are flapping once or twice in an hour, you can open it as a ticket in a Service Now system. But if it’s like excessive flapping, then you can immediately shut down that interface. So we can introduce automatic remediation that works across multiple vendors using a workflow engine, and we can also integrate with tools like Grafana so we can visualize all the metrics and generate reports and compliance reports as well.
This is a bit of detailed slide. It tries to explain how we achieved this. The ATOM platform is composed of many microservices. All of these are Docker containers. They provide individual functionality like discovering the devices, building the topology, collecting the telemetry data, creating a service chain across multiple vendor devices as well as delivering analytics and visualization of the network infrastructure. We integrate with 45 different vendors – all the leading vendors: Cisco, Arista, F5, Citrix, Juniper – you name it. We have it. All the networking related devices we can configure either using CLI, NETCONF, API or REST as well. And all of this data is stored in a time series database with which you can query their data and generate all kinds of reports. And we use kubernettes to manage the various Docker containers within the ATOM platform, so these can be deployed on-prem, or these can be deployed in public clouds like AWS and Azure as well.
So here is the customer deployment – as you can see it’s a fully redundant design. They have multiple data centers – The two instances of the ATOM are synchronizing with each other. The policy data, as well as the times series data, is constantly synchronized so that you have a fully redundant design. And we will go through a short demo of how this whole integration looks like.
Device Onboarding and Topology
So let me switch to the demo. This is a recorded demo so I’ll try to pause it so you can follow along. So first we are going to show a dashboard of all the devices. We then discovered the devices; we can see the various devices, and the config and the ATOM is sucking up all the configuration data from all these vendors, and it normalizes it into a JSON object. So now it builds a topology of your network – it creates this concept of resource pools, and you can look into details of each of their devices. As you can see it overlays operational data on top of this topology. So that gives you full visibility into your entire geographical distribution.
Configure Pre-defined Telemetry Sensors
And then we are configuring telemetry data, the sensors – on this case we are configuring the interface statistics metric on the Cisco IOS-XR. You can decide exactly which data you want from this particular IOS. We are looking for interface status stats, and we can push this config across hundreds of devices. Now from a single GUI, you are configuring the Arista, the Cisco, the Juniper and ATOM will go through a workflow engine. It’s going to show you now the model that it is going to push to the device. As an operator, you can look at the model and verify that everything is working correctly and you can approve it and schedule to push this config at any particular time.
So once it is approved the ATOM is going to auto-generate the commands for various devices whether it’s the Cisco or Arista or Juniper as you can see it’s going to show you the Cisco CLI that it is going to push to the devices. Now the devices start sending that telemetry data to the ATOM platform. And you can see the data in the Grafana dashboard. So here we collect various metrics like interface Utilization and CPU utilization, interface drops and things like that – the dashboard is highly customizable, you can build your own reports and have different alerts created out of it.
And because this customer is so particular about the latency, we have collected the latency within the ATOM platform. You can see that in 26 milliseconds we are able to take the network state like something happens on the network and that data will be stored in the database for applications to access within 25 milliseconds. That’s the kind of latency we are dealing with. The ATOM platform optimizes the overall end to end latency by using the latest software stack, and it works across multiple domains. So it’s a really short demo.
The main takeaway is that the ATOM platform is able to meet the latency requirements and help the customer deliver very fast and responsive networks that are working across multiple vendors and that has a very fully redundant design using open standards.
So here is the final slide – as you can see, the ATOM scales to millions of devices using micro services-based architecture. It introduces closed-loop automation that delivers analytics, remediation, monitoring as well as orchestration for multi-vendor infrastructure. And it also can be deployed on-prem or in the public cloud like AWS and Azure.
You can check out the website Anuta Networks for more case studies as well as datasheets. Thank you.
