AT&T has made some waves in the industry with its ECOMP software management architecture but does it go far enough to address the challenges of cloud native VNFs? ECOMP is basically an attempt for AT&T to drive more automation into its operations. And AT&T has set some very aggressive goals of 75% virtualization of its infrastructure by 2020. ECOMP is intended by AT&T to be “more” than MANO, because its intent is to deliver a high level of end-to-end, cross-domain orchestration and therefore transform AT&T businesses and cost structure. However, while ECOMP is moving toward that end-to-end view, we have seen little that suggests that AT&T and its suppliers have made significant progress on re-architecting and de-composing VNFs to be truly cloud-native. We believe this is a critical step, because higher availability, better capex utilization and greater automation are driven from principles of cloud-native that allow the platform to optimize itself. This increases capacity utilization, allows for proactive healing, and entirely takes humans and higher-level orchestration out of significant segments of the lifecycle process.
ECOMP orchestration is more advanced than simple VNF instantiation because it factors in integration of legacy OSS/BSS which will be around for a long time. But today, it must orchestrate sub-optimal NFs – often large, pre-engineered configurations that look suspiciously like their pre-virtualized selves. Witness that several companies sell “vEPC” – a large aggregation of NFs! This is totally unlike the operations of web scale companies, in which micro-services within each NF scale, heal, and load balance across datacenters and across the globe automatically, in the background. Cloud native in the context of telco networks is the ability to deploy services in seconds with a service lifecycle that may only last in timeframes of minutes or hours. This is a big leap from services offered over a physical network that take months to deploy and where service lifecycles are measured in years!
Recent announcements from AT&T suggest that the 75% virtualization of its network will extend far beyond 2020 and this supports our forecast that we published in 2015
https://appledoreresearch.com/report/virtual-network-capex-global-forecast-2015-2025/
Virtualization of a network function on a router, firewall, or elements in the mobile packet core have already been proven in live deployments. The challenge is how do you deliver full life-cycle management across many domains – some of which span outside your sphere of control.
Cloud native applications combined with virtualization of the telecommunication network has the potential to substantially improve capital efficiency for CSPs in an industry that spends $300 Billion annually on network, compute, and storage. Currently CSPs are actively virtualizing network functions in the mobile core network, customer premise, and data center to both lower capex but more importantly change the operating cost curve that exist today in their operating environment. The cloud enables CSPs the ability to take advantage of delivering services more rapidly to their customers at a far lower operational unit cost.
We believe that this journey towards virtualization which combines NFV, SDN, and cloud native applications will be more disruptive than the move from circuit switched to packet networking. The cloud native approach is application centric where applications consume resources, scale based on customer demand and are completely de-coupled from the infrastructure. Cloud native applications utilize micro-services based architecture not a monolithic software stack which is pervasive in every CSP operational environment today. The challenge for CSPs is how to rapidly transform their processes and systems to achieve the inherent benefits offered by the cloud and virtualization. CSPs have systematically planned, designed, and deployed services on a network infrastructure assuring five 9’s of reliability. Unfortunately, it has been a network centric approach that dictates how and when services will be instantiated, provisioned, and managed. The tight coupling of the network to service deployments combined with a monolithic management stack inhibits CSPs from fully realizing cloud native virtualized network functions.
To speak with us at MWC Barcelona on the architectural guidelines in the transformation towards cloud native or schedule a one on one session remotely, inbox me at Patrick.kelly@appledorerg.com