21- 23 May 2019   Hotel Palace,  Berlin

Please outline the aims of your network virtualization case study: 

TURKCELL is a digital operator headquartered in Turkey, serving its customers with its unique portfolio of digital services along with voice, messaging, data and IPTV services on its mobile and fixed networks. TURKCELL has created the Unified Telco Cloud to implement NFV infrastructure and centralize that Core Network for Telco and Gi-lan services. The Unified Telco Cloud is preparing TURKCELL for 5G and edge computing evolution. It’s goal is to lower the barrier of entry to our own network to make it a breeding ground for innovation and competition while unifying the infrastructure and decreasing operational costs. One significant aim towards building the NFVI is the organizational differentiation whereby each layer of Telco Cloud is designed and managed by TURKCELL Core Network Units and it focuses on the ability of processing data plane services with high throughput and high pps requirements. TURKCELL has set out with challenging services such as Deep Packet Inspection for traffic management and enhanced VAS, Carrier Grade NAT for IP address management, Data Optimization for differentiated quality of experience for customers, and also control planes services such as VoLTE PCRF, Mobile PCRF, Mobile AAA for policy and charging, Enhanced Communication Services etc.

What are your KPI's?

TURKCELL has deployed one of the largest virtualized mobile networks in the EMEA region, supporting more than 35 million mobile broadband subscribers, 99 percent population coverage and carrying over a terabit per second of customer traffic across the region. TURKCELL NFVI platform is capable processing up to 1.8 Tbps of traffic. TURKCELL has now virtualized 18 percent of voice and data services capacity on its telco cloud percent. By the end of 2019, TURKCELL is planning to virtualize 40% of its Network Services.

Outline all of the challenges you face:

TURKCELL has started its virtualization journey by virtualizing its major data plane services that has been one of main challenges. Unlike IT services, to get high throughput and high packet per second, Unified Telco Cloud infrastructure uses complex packet processing technologies such as DPDK enabled OVS and SR-IOV. It focuses on CPU level optimizations such as CPU pinning and NUMA awareness. Each server has a PCI design that allows multiple NICs to be processed by separate CPU sockets in order to get the highest CPU and RAM performance. These improvements provided horizontal scalability and service agility capabilities to help maximize the efficiency and effectiveness. In the Unified Telco Cloud, unifying services with different underlying network designs on the same virtual infrastructure and exceeding the capacity provided by the legacy world were two of major motivations.

What partners will you be using to help overcome these challenges?

TURKCELL’s centralized cloud utilizes Red Hat OpenStack Platform, the carrier-grade distribution of the open source, massively scalable infrastructure framework for building hybrid clouds. One of the major challenges TURKCELL has come across during this transformation is working and collaborating with more than ten vendors, which requires significant coordination effort. From NFVI and service tenders to production stages, TURKCELL focused on the NFV specifications described in the ETSI Reference Model and displayed a solid stance that also motivated each dealer. This behaviour is one of the major contributions of TURKCELL to the maturing NFV standardization. Each element in the NFVI was considered referencing ETSI reference architecture. Community contributions and vendor independency are two major constraints TURKCELL has followed during the evaluation and implementation of services.

What makes this case study different?

The Unified Telco Cloud platform’s variability provides TURKCELL with industry leading time-to-market for new service introduction, significant capex savings, and is one major example for converged Fixed and Mobile Services. We adopted innovative approaches in operations through the use of multi-layer alarm correlation by adapting various open source tools. This integration provided the root-cause analysis capability, a major requirement for end-to-end NFV environments. Systems that previously operated over separate network topologies converged over a single network design. Enabled different nodes to interoperate over the same NFV infrastructure. New structures like scale-out solutions and load balancing algorithms provided for the required capacity.