Discussion on Applying a Converged CDN in the Access Office

 

 

Status and Challenges of CDN 

Although the internet is ubiquitous now in people's life and work after decades of high-speed development, content access has always been one of its primary functions. Increases in content types, video bitrates and terminal types all lead to rises in content traffic. Consequently, the bottleneck of core networks becomes more severe and degrades user experience. This gives rise to a content delivery network (CDN) that can improve quality of experience (QoE), reduce backbone network traffic and properly distribute content traffic. By directing user requests to suitable service nodes through content delivery and service scheduling, CDN provides distributed content service. 

Depending on the content they deliver, CDNs fall into two categories: specialized CDNs and converged CDNs. A specialized CDN delivers a specific type of content. For example, an IPTV CDN specifically carries IPTV video content. A converged CDN can deliver multiple types of content and support many different terminal types. Operators mostly choose the converged CDN in their construction.

 

Currently operators run a converged CDN at the provincial level. The CDN employs a two-level architecture that comprises a provincial-level central node and city-level edge nodes (Fig. 1). In few large cities that include high-traffic districts, the edge nodes have moved closer to the broadband network gateway/service router (BNG/SR) level, forming a three-level CDN architecture.

The above architecture has both advantages and disadvantages. It can significantly cut backbone network traffic and long-distance (from prefecture-level city to provincial capital) transmission cost. Nodes at both levels can be massively deployed with a mature construction solution. The nodes can be co-located with edge and provincial data centers (DCs) to share a computing and storage infrastructure. However, the architecture also has obvious disadvantages. First, edge nodes are placed far from end users. As a result, many hops are needed in a service path, QoE is difficult to assure, and fault location and removal is complex. Second, the network segments from OLT to BNG and from BNG to boundary router (BR) tend to become bottlenecks as service traffic surges. Finally, this architecture does not leverage the low-cost and abundant-bandwidth advantages of an optical access network. The disadvantages will become ever more noticeable as 4K/8K/AR/VR video content become ubiquitous.  
With the emergence of edge computing and the introduction of SDN and NFV, it has become a trend to introduce NFV infrastructure (NFVI) in the access office. If the NFVI in the access office can be fully utilized so that converged CDN services can be deployed near end users and a three-level CDN architecture can be built, the disadvantages mentioned above in the two-level CDN architecture can be effectively avoided. This article is intended to discuss this approach. By introducing a new converged CDN solution based on ZTE’s TITAN platform, the article gives good thought and suggestions for operators to deploy converged CDNs in their access offices.

Introducing A-CDN to Build a Three-Level Converged CDN Architecture

An access CDN node (A-CDN) can be introduced in the access office to form a three-level converged CDN architecture together with existing edge nodes and a central node (Fig. 2).
An optical access network based on the PON technology provides high access bandwidth capable of supporting heavy-traffic services including 4K/8K TV and AR/VR applications. As shown in Table 1, in the FTTH mode with a split ratio of 1:64 and a concurrency rate of 40%, GPON can support 4K TV while 10G PON support 8K TV and AR/VR commercial deployment.

 

Nevertheless, a bottleneck exists in the uplink bandwidth from OLT to BNG. Fiber resources between OLT and BNG are limited. An OLT has no more than four pairs of uplink fibers and each pair usually offers 10GE bandwidth, meaning that an OLT can provide only 40GE uplink bandwidth. An OLT carries 5,000 to 8,000 broadband users and busy-hour traffic per household is around 3 Mbps, which translates into a total traffic of 15–20 Gbps. Assuming that the utilization of uplink bandwidth is 70%, an OLT can carry 28 Gbps traffic in total, which means a bottleneck in the uplink bandwidth is imminent. 
This bottleneck of OLT can be addressed in two ways: by increasing uplink bandwidth or moving the CDN closer to end users. The uplink bandwidth of OLT can be increased by either upgrading the uplink port or deploying OTN to the access office. Upgrading the uplink port from 10GE to 100GE requires the support of both OLT and BNG, which results in high costs and a long period of time. The upstream equipment also needs to be upgraded. Deploying OTN to the access office involves even more engineering work and higher costs. However, moving CDN closer to end users can quickly remove the bottleneck in a flexible and cost-effective manner.
The development of IT technology follows Moore’s Law, which means that IT equipment cost decreases far faster than network equipment cost. As CPU capability increases dramatically and its memory and hardware performance develop rapidly, a common server can support a CDN processing capacity of dozens of gigabits per second. Consequently, it is economically feasible to get more network bandwidth at the cost of computing and storage. Meanwhile, the FTTx construction and the increasing port density of access equipment are releasing the frame and rack space in an access office. By embedding blade servers into access equipment, A-CDN can be built quickly and flexibly as desired without the need to reconstruct the access office.  
An in-built A-CON is achieved by embedding CDN functionalities in OLT. Because a large-scale content storage can not be configured, the content scheduling algorithm of CDN needs to be optimized. In other words, A-CDN needs to store the hottest programs to improve the content hit ratio, and users can access services from the nearest A-CDN.  

ZTE's A-CDN Solution

ZTE has developed blade servers that are embedded in its TITAN new-generation OLT platform to provide an A-CDN solution.

In hardware design, the blade servers employ a powerful SoC-based CPU, NVMe SSD drives, and 10G network ports. Each blade server occupies two service slots and can be inserted in any slot in the TITAN OLT. In software design, the embedded blade servers run on ZTE’s proprietary CDN software. Evaluation and verification show that the blade servers can deliver a converged CDN service capacity of 20 Gbps and provide 70% of the CDN service for 20,000 users in the same access office. This vastly reduces OLT’s uplink bandwidth of the converged and metro networks and delays the need for chain network upgrade. To suit the A-CDN scenario, ZTE's CDN software system has been optimized in storage and content.

● A-CDN storage optimization: Store three types of most broadcast content including three-hour live-channel time-shifted TV programs, TVOD programs in  a week, and 5,000 movies that are currently broadcast most. A-CDN has a total storage capacity of less than 16 Tbps but can cover 70% of the VOD service.
● A-CDN content update optimization: Use big data to analyze and optimize regional hot content and the cooling model of hot content so that the cached content can be promptly and accurately updated and scheduled.
TITAN has embedded the traffic offload functionality to avoid adversely affecting VLAN, IP planning and BNG user management. Through listening and learning, TITAN offloads according to destination IP addresses the traffic generated to access A-CDN. It implements PPPoE/IPoE gateway functionality in advance without affecting the other traffic. This mechanism is fully transparent to BNG/SR, STB and A-CDN, which facilitates engineering and deployment. 

Conclusion

As video traffic grows exponentially, it is increasingly important for CDN to ensure user experience. The A-CDN architecture and key technologies are used to embed CDN functionalities in OLT. The low-cost flexible A-CDN solution can help operators rapidly upgrade capacity and deploy their big video services to meet future traffic challenges.