As automation-driven technologies such as big data analytics, machine learning, AI and other autonomous solutions gain ground, network demands and expectations are increasing manifold. At the same time, moving away from legacy technologies and into enterprise cloud solutions requires automation. And this entire cycle is being driven by 5G.
As the Linux Foundation's general manager for networking and orchestration, Arpit Joshipura is tasked with solving network-operator conundrums like this through cohesive open source solutions.
Last month, Telco Transformation interviewed Joshipura on enterprise cloud from an open-source perspective. (See Linux Foundation's Joshipura on Building Open Enterprise Cloud.) Now, in Part II of this Q&A, Joshipura delves into the inter-relationship between enterprise cloud, big data, automation, and 5G networking.
Telco Transformation: What is the top challenge you have identified for enterprise networks that may still be too enmeshed in central IT?
Arpit Joshipura: The number-one challenge is automation. Automation starts from a design concept and goes all the way to pushing that design to the infrastructure, and then actually running it in operations and upgrading. It's the entire lifecycle automation of a service that an end user pushes out … You have a double advantage there where you not only speed things up but you also move and retrain your staff towards next-generation services.
The problem of automation is that it's critical for the network operators, more than for even the enterprise, and the reason is 5G. The requirements for 5G are very clear. You've got higher throughput. You've got lower latency. You've got billions of connected devices --or IoT -- that will have a lifecycle need for automation on these networks. So automation is mandatory for 5G, and now's the time to automate before 5G hits.
If you think through a very simple use case, a connected device is not going to be placed on hold while an operator answers the question of "Do you want to come on the network? Let me bill you. Let me charge you. Let me add you to the server." It's not happening. These things come on the network in milliseconds, stay alive, the entire lifecycle could be minutes, and then they sleep for a week. So it's a whole new dimension.
TT: Tell me more about the role of automation, AI, and real-time predictive and prescriptive analytics in the 5G evolution.
AJ: Okay. These are workload/apps on 5G, so we have to really look at the role of 5G to enable this whole ecosystem of AI and predictive analysis and data … What 5G technology brings is more bandwidth, lower latency, a lot more devices, and a lot more connectivity options, but what it is not bringing is a layer of automation … What we are doing in open source is we are bringing that layer of automation so that these industries or these applications can flourish because they need -- beyond just the 5G bandwidth and lower latency -- a layer of automation to do the analysis and to move data back and forth and to connect these devices that actually do analytics, without human intervention.
TT: What are some of the projects that the Linux Foundation is working on in terms of enabling that layer of automation?
AJ: The most important one that we just recently launched is ONAP. ONAP (Open Network Automation Platform) is a merger of two open source projects. One was OPEN-O, which was kind of led by China Mobile Ltd. (NYSE: CHL) The other was ECOMP (Enhanced Control, Orchestration, Management & Policy), which was led by AT&T Inc. (NYSE: T) along with BCE Inc. (Bell Canada) (NYSE/Toronto: BCE) and Orange (NYSE: FTE). We announced at Mobile World Congress a couple of weeks ago a merger and a bringing together of the two communities, where the focus is on automation. It is automating everything from the control plane up into orchestration, into the analytics, into the policy engines, all the way up to the design and inventory control -- so that when 5G services and applications like analytics or AI or predictive analytics come on board, they can utilize this layer of automation. ONAP has gotten tremendous response and support globally. More details will be announced at ONS (Open Networking Summit 2017) on that project.
TT: How do you see the role of the evolving enterprise cloud in supporting the AI and automation demands of the future?
AJ: So in a typical data center or in a typical enterprise, historically, when an employee wants to run a traditional Internet workload, the typical traffic patterns are what is called north-south -- meaning traffic goes from the data center out into the Internet, and into either another data center or SaaS or any "-as-a-service" kind of service. When you are architecting a workload for a data center that has a typical north-south pattern, you would architect the data center with traditional technologies that have been around for the last 20 years.
The new modern workloads -- whether it's big data, Hadoop, analytics, or even some of the more modern application software -- mimic web-scale workloads. They create a lot of traffic that is east-west across the different server and the data center racks -- not north-south. It is not from the] data center out into the Internet, but within the data center.
Coming back to the analytics and the AI and a lot of the new modern workloads, the big role of cloud is to make sure they architect for east-west traffic patterns. It is a really flat data center network. It is built with a way to interconnect these servers so that it is optimized for services like big data and Hadoop and these modern workloads. What it effectively does is it allows you to talk a lot more internally.
Both hybrid and public cloud have to be evolved in relation to a modern workload like analytics or like AI because people think that when you have to do AI, big data, whatever, all you need is bandwidth, and bandwidth is just not enough. Yes, 10-Gig, 40-Gig, 100-Gig of throughput is coming to the data center and has been deployed, but it's not just about bandwidth. It's also about traffic patterns. And then with 5G in the network it's all about latency as well.
— Joe Stanganelli, Contributing Writer Telco Transformation