 Hello everyone. Although the life of telco began over a century ago and wireless networks are about four decades old, the virtualization of networks and orchestration of network function is relatively a new concept. And yet the world around us, undoubtedly today hinges on communications networks. The concept of cloud computing is a relatively new one. Most of us, if I may say, remember the time when cloud meant only one thing and the cloud computing didn't mean anything. And today, today nothing really happens in the internet without a cloud. I would like you to imagine if these two technologies, what if the evolution of these two technologies influence each other more than a mere coincidence or afterthoughts. This is the key theme of my today's talk. In the common parlance, the word orchestration means anything between something that enables service lifecycle management or virtual network function lifecycle management, automation of network and service configuration or simply something that manages network complexity through automation. In the rest of the presentation, I would end up using perhaps every one of these definitions and a few more when I use the term orchestration. Here is the agenda for our talk. First, we will look at the orchestration through different angles and then double click on how the co-evolution of telco and the cloud technology is taking place and how it could be slightly different. And then touch upon a very important topic of the need for advanced algorithms in this space and conclude our conversations with a few takeaways. Here are the four main facets of orchestration ecosystem. The first one is about multiplicity. It is about having multiple network functions, consequently having multiple vendors, partners, multiple systems, and even the vendors for orchestration tools and then the adjacent OSS and the BSS system. The multiple vendors of the partners for network functions brings forth the role of standards and compliance to bring uniformity and simplicity for the operations. When it comes to orchestration tools, they are again driven by standards such as HC, ZSM, parts of 3GPP, open source like ONAP, and partially by other standards bodies such as ORAN and MEF. Furthermore, how the orchestration tool chain touches the OSS-BSS layer is driven by TM Forum. The next important facets is all about the network domains. The obvious one is 3G to 4G to 5G and corresponding network functions and topologies. But remember, orchestration applies in equal measure and forced to wireline applications. This is an integral part of industry 4.0 transformation. The second angle is look at the opportunities and the requirements. Look through the angle of end user. End user or end applications require in some cases network ubiquity, rich applications, services requiring high bandwidth, low latency, a different kind of reliability or even intelligence. Operations and business on the other hand is dealing with multiple clouds, multiple layers of the infrastructure, multiple domains, operational sustainability and end-to-end need to deployment and automation. And the business on the other hand looks at the 5G characteristics and the services which are based upon these 5G characteristics. Network elements needs to be distributed in some cases in order to give you those characteristics and scalable and flexible. The one more way to look at this particular facets is to look at the industry angle, quickening of the pace in standards. There are places where standards are progressing really very fast. There are multiple standards bodies, 3GPP to HC to GSMA to ITF to math and everything in between. And at times there's a good convergence and at times there's a slow convergence and I have to report to you that even in some cases there seems to be a divergence. And lately the focus on open source even standardization of testing and test beds has also intensified. Finally, there's a renewed focus from government and research and development R&D bodies in this particular space. Cloud infrastructure is a transition from legacy infrastructure to contemporary which is general, most cases general purpose, but in quite a few number of cases it is purpose-built hardware, monolithic application to cloud modernized or and the reorientation of a core and packet IMS requires network as a service, analytics as a service, security as a service, anything as a service. In order to meet the capacity requirement cloud infrastructure requires multiple clouds, multiple layer platform to host heterogeneous application to offer variety of services. Two important subsets of this facet is private and public. Similarly, the nature of workload also influences the architecture and composition of the cloud. And this is the point that we will discuss further in a few minutes. Network function on the other end is a cocktail of services running as a network function ranging from network functions implemented as simple VMs or cloud native realizations or cross-skis cases such as VNFs running on containers or CNFs running on VMs and having a diverse SLA requirements that are hosted on the shared capacity pool of cloud infrastructure. All aspects of management and orchestration from package structure, uniformity to the life cycle management to the composition of network function for slices are the key constituents of this particular facets. So what is the goal of orchestration? It is to transform network to deliver high quality and cost effective product and services. Some really seen implication of this are the network automations, agility of service realization, flexible control over SLA management, economies of scale, vertical integration for better customer experiences and much more. And what are the means to these meeting these goals? The very first one is standardizations. So traditionally there has been a very good focus and progress on interface and profile standardization led by 3GPP, GSMA and quite a few number of other standards bodies. There has been a lack of equivalent progress when it comes to data model standardizations though. An example is much needs to be done to standardize the models in the management plane. The next important tool in this box is simplifications, right? Take an example of a service design to meet the telcos unique need. An example would be to exploit the full potential of Kubernetes. One would want to design the flow once and exercise it for multiple CNS and also continuously reduce the number of steps in this process. There is never enough automations. Automation is something that needs to be done day in and day out. There are three, primarily there are three steps here or phases here. The first one is pre-insanctations which is as simple as security compliance, standard validations, automatically identifying the right places in insanctations and preparing it, moving the binaries there. The next one is insanctations. So this is the insanctation of network functions. It could be a service chain, it could be a slice through VM-based or Kubernetes-based or any other related platform technology. And then immediately performing the network configurations to properly embed this newly constructed entity into the underlying networking fabric, typically L2-L3 fabric. The next part and there's a post-insanctation, once post-insanctations or insanctation which is to configure the application settings and policies or even updating the other applications and their policies to essentially make the newly insanctiated entity operational. And this process requires, this whole automation process requires continuous revisitations and the design and deployment. The last and a very critical means for meeting the orchestration goal is the cloudification of applications and network infrastructure. Starting with the early clouds and accentuated with the modern clouds for deploying the network functions across the full breadth of telco from RAM to EPC, wire line to wireless with varying degree of resource requirements. Some are database intensive, others are latency sensitive and many are computationally intensive. At times, bringing the NFE workload into the cloud is like mapping a square peg into a round hole. And I say this because of the available vectors such as abstract complexity due to the gaps in the standards, nick assignment, auto-scaling, auto-hailing, lack of upgrade strategy supported in the open stacks and lots of restriction imposed by the cloud itself in case of what is allowed and what is not allowed. Cloud evolution have roughly speaking three phases. The first phase was all about or we can characterize the pre-cloud which is all about small data centers, limited scales and heterogeneous environment. The early cloud or the second phase was all about huge data centers and infrastructure as a business or better known as infrastructure as a service. Today's cloud are entirely different from the first two categories. These are purpose built hardware and software optimized for multiple use cases, vertical integration delivers extreme optimizations and very customer user experience. And the hallmark of today's cloud is physical infrastructure is built from ground up unrestrained capacity scale clause, virtualized networking and multi-tenancy. Vertical integration is about OSN hardware to optimize the memory, the CPU and the disk resources. Also use of hardware integrated with artificial intelligence services, even intelligent NIC and so on. The third and the most important thing has been the focus on telemetry, the design on testing or testability, AI driven operations, having the placement decision being made smartly. And finally, security as an integral part of this thing in an environment that allows multi-tenancy, the focus on security is a must. Zero trust models, data masking, encryption, securing the server and quite a few of these security related technologies. And what comes next? Well, it could be anything from neuro morphic chips to the holy grail of computing, which is a quantum computing and many things in between. When you look at the trajectory of the telco transformation, on one end it looks as a very simple one, which is focus on a voice, got to the focus of the texting, then focus on the internet connections. And what you see next is the 5G, which is a revolution unfolding right in front in all of us. The other way to look at this transformation is to see how the use cases have evolved and the various different needs have evolved. So one of the things is look at the traffic pattern and how they have dramatically changed both in terms of the volume, the distribution of the volume, and the source and destination of the volumes too, right? Everything seems to be drastically drastically have changed and have been changing. The other thing is complex industrial use cases, whether in the context of IoT, in the context of AR, VR, in the context of all other kinds of advanced technologies that has been changing. The, there has been a phenomenal growth in the radio and the transport technologies. At the same time, you're moving towards in a more spectrum, better spectrum, and better combination of more, you know, the spectrum. The open standards, the standards of the open sources have been the key enablers for this particular transformation stuff. Ability to create cost effective and high quality experiences for our customers require a very high degree of network automation. And in that sense, network automation is the necessary ingredient for any and all practical forthcoming network slices. This transformation will be incomplete because of the mention of the security which is at the heart and soul of it. Next, let's look at the projection of cloud technologies on the telco needs. As the cloud is moving towards the edge, more brain power is needed at the edge to reduce the backhaul network traffic when compute and storage nodes are closer to the proximity to the users. This requires the cloud platform to be more nimble and faster. The quest for space efficiency and energy efficiency as a direct consequence of going after cost effective and high quality solutions and services is never ending. Besides reducing power consumption is simply the right thing to do for the environment. There have been a major advancements in cloud technologies which now support multiple new runtime and compute engines such as dockers, serverless, unicornals and so on. The pace of adoption of these emerging tool set in the telecom industry admittedly is slow due to the lack of support of the NFE standards and simple beginning. Multiple technologies, characteristics of NFS give rise to heterogeneous application such as communication intensive, computational intensive or the storage intensive. They pose unique requirements such as various flavors of cloud networking, security along with the programmable infrastructure to telcos especially for the edge use cases. Some functions are deployed as individual resources and have no dependency on other network functions. Whereas some network functions define mobile services such as 5G network functions are often deployed in the mesh, they are interconnected and some edge applications need hyper-connected computing environments. Such complex mesh network requires interweaving of cloud and network top to gain access to the advanced control and the data plan to perform advanced operations such as service discovery and network observability, analyzing the traffic trends, health checkups, health checks of the services spanned across multiple domains. In order to meet the low latency and high throughput requirements, the networking fabric has to be quite sensitive to these needs and the same is true about the forwarding plan. Confidential computing has been a major breakthrough technology in recent years to preserve the confidentiality of the data when in use. These can benefit the advanced use cases for telcos such as network slicing for achieving greater level of isolation. The priority slices carrying highly sensitive information of users, governments can be isolated from the other slices and can be secured in what is called the secure enclaves in the hardware and inaccessible to untested code. Finally, telcos are happy to note that standards and open source communities of cloud technologies are adding telco specific streams and initiatives. An example is a telecom user group of CNCF or OP NFE and the CNTT coming together to form new focused initiatives. Next slide, please. To unleash the full potential of 5G revolution, the customer has to be able to enjoy the new product and services with the superior service characteristics such as latency, high throughput, super energy efficiency, device density and quite a few of those. Be this is for a simple smartphone or a mission critical IoT device or just a fixed wireless access. So therefore it is imperative that cloud technologies to fully consider these in their evolution plans. So this is the first of the many expectations that telco has from the cloud technologies. Telco has some of the most stringent needs when it comes to the data plane. Data has to be processed quickly, efficiently supporting both the high throughput and low latency. This requires cloud technologies to imbibe high performance processing and acceleration methods such as DPDK, SRIOV and quite a few of these technologies natively. Service aware routing mesh networking and 5G where applications are interconnected with each other as per 3GPP defined SBA, telco sees an opportunities in the development of service mesh technologies to perform service discovery of applications which are either geographically co-located or distributed. Depending on the real time workload, service aware routing algorithm can provide optimal path to dynamically steer the traffic, including a framework to do so based on the slicing priorities. The fourth expectation is opportunities in the hierarchical layer of the management plane, NFE, MANO and the control plane, Kubernetes, OpenStack and so on. Multitenancy means the same infrastructure and control plane is shared by the multiple users and entities. Opportunities are designed of the new tendency model to slice the application traffic to isolate from trusted and non-trusted or less trusted tenants. And at the same time, tenancy can be achieved at the management plane to slice the network elements and prioritize the traffic based on the interest slice or interest-lice priorities defined in the NSTs and so on and resource isolation right at the hardware and the software level. The number of network functions virtualized for the first time is steadily increasing. And the number of instances of network functions that are getting instantiated is increasing exponentially. Therefore, it is expected that all telco specific extension to the cloud technologies are robust, fully automated, scalable and to support, you know, there's pets to cattle transformation that is very much needed now. A sophisticated telco network servicing the needs of a diverse customer based and with the, you know, quite a few of experiences has multiple layers and segments. The nature of the data and the volume of the data, the time scale of decision-making in each of these layers and segments is different. Yet a behavior or policy influence in one can easily impact the other. So while the need for local optimization techniques cannot be dispensed with, the quest for global optimization and decision-making has only begun. Now that requires close observability at any given time how the network is behaving and why. More importantly, more work needs to be done with full assistance from the cloud technologies is to achieve this without impacting the network itself. Now let's look at the next theme which is about advanced algorithms, right? In the interest of time, I'm just going to touch upon a couple of these. So use of AI ML techniques to reduce energy consumption by turning network elements on and off preemptively rather than waiting for the traffic levels to change has become extremely, extremely useful. There has been a steady increase in the overall dimensionality of this problem. It is no more just about power or space or cooling or such single dimension objectives, continuously varying traffic patterns, mobility patterns adds another layer, not just of the complexity, but also an opportunity to squeeze more out of the solution. The next I would touch upon is the application placement, right? Operators usually rely on multiple workaround techniques such as manual or static or even pre-configured, scheduling policies, duplication of resources geographically co-located or distributed sites to meet application or service level requirements and business SLA requirements. Dynamic placement and optimization in the dynamically changing environment is essential. With onset of 5G has a significant role in the success of 5G itself. 5G networks therefore are designed to be tailored or sliced to meet the telco specific use cases and end user requirements. Advanced algorithms using AIML could frame the placement policies to apply conditional automation logic to perform semi-autonomous or fully autonomous placement where necessary to maintain the quality of service requirements of that particular application. Okay, now let's tie all this thing together. It is absolutely a must to have a singular focus on operations while architecting or implementing orchestration solutions. Success of orchestration is measured through network automation in its ability to reduce the complexity. Network automation is the strategic imperative implied by the orchestration. The link between these two is so strong that they are almost interchangeable. One logical consequence of this is cross-domain network orchestration which is a fusion of NFE, the network function virtualization orchestration and the software defined control and automation of the van. Similarly, one has to look at the multiple clouds as the need vector is different depending on where the cloud is located on the telco's network topology. As orchestration becomes more and more mainstream the network automation extends both its reach. There will be a need to use full set of tools from basic automation through APIs to the real-time sensing of the network to AI-driven closed loops and much more. What began as a pets are today's cattle. While each network function has its own distinct role and each vendor has its own niche to keep the orchestration pragmatic there has to be a layer that is uniform for service life cycle management for exploiting the cloud technologies to the fullest. We believe standard is an effective tool to keep the network simple. So pursuing it diligently and pragmatically is the way to move forward. Finally, there is much that network cloud technologies could offer to telco to continuously offer our customer with newer and better experiences. And telco can shape the evolution of the cloud technologies by offering its unique needs. Thank you very much.