Simple AO: The Road to Access Office Re-Architecture

Nowadays, the telecommunications industry is experiencing massive changes. New generations of technologies, new concepts, and new services keep attracting people. 

 

One of the hottest topics for the community is 5G, which is expected to enable lots of exciting services including ultra-high broadband, ultra-reliable and low-latency communications, and massive connections. Comparing to 4G, 5G has more value-added services and more expansions, but it brings more new challenges to the network, such as 100 times transmission rate, 100 times number of connection terminals, and an end-to-end latency of less than 1ms.

 

Many operators are participating in the 5G arena. All the top operators in the United States, China, and Japan, the world’s three largest economy bodies, have proposed their plans of the commercial deployment of 5G networks near 2020, and more countries are keeping up with the trend. 

 

Besides that, Software Defined Networks (SDN) and Network Functions Virtualization (NFV) are another popular and frequently talked-about theme in the industry. Many tier 1 operators in North America, Europe, and Asia-Pacific are quite interested in SDN/NFV since they expect such new ideas are capable of helping reduce network construction costs. 

 

For the operators focusing on fixed-networks, the emerging services of big video is a good news and is helpful to raise revenues and competitiveness. The industry has reached a consensus that the video services would occupy most of the Internet traffics. Therefore, the big video services would definitely be a plus for the traditional high-speed Internet and voice services.

 

In the 5G era which is full of big changes and innovations, as the node is close to the end users in the fixed network, the Access Office (AO) needs to be re-architected to satisfy the requirements of future networks. Considering the popularity of mobile communications and the virtualization technologies, it is believed that the thoughts of fixed-mobile convergence, IT-cloud convergence, and multi-scenario support would be an inevitable trend.

 

Specifically speaking, we propose that the AO in the future should include OLTs working as the converged optical node. Moreover, such OLTs should consist of two major parts, integrated Physical Network Functions (PNF) and light cloud. The PNF brings the AO the capability of different types of connections. While the light cloud brings the AO infrastructure, based on which multiple virtual network functions and storage purposes could be realized.

 

Normally, the AO connects terminals with different user types like residential, enterprise or government, and mobile base stations. Since different kinds of terminals require different connection features, including bandwidth, latency, security and reliability, various access network functions co-exist in the same AO. For example, XG-PON or XGS-PON is suitable for residential users, while P2P is often chosen for enterprise or government users, and mobile base stations. However, in 5G era, the number of 5G AAU would be extended rapidly by approximately over 10 times. That means a large amount of fiber resources are required if we keep using P2P as fronthaul. In some scenarios like the deployment in dense residential areas, the existing FTTH network brings abundant fiber resources and reusing them for 5G fronthaul could be a cost-effective choice.

 

WDM-PON is a good candidate for 5G fronthaul. Firstly, as a wavelength-division-based technology, WDM-PON avoids the delay caused by the time slot assignment mechanism, which is commonly used in time-division-based technologies such as XG-PON or XGS-PON. According to the estimation, when using WDM-PON together with some latency control mechanisms such as TDM-like channel, will reduce 59% end-to-end latency, and meet the uRLLC requirements for 5G fronthaul. 

 

Secondly, WDM-PON has the same P2MP network topology with the FTTH network. That means the operators do not need to re-design or reconstruct their access networks. More importantly, unlike the P2P network, the P2MP network contains a “trunk” segment and a “feeder” segment. Therefore, multiple 5G AAUs correspond to multiple feeders, but share the same trunk, which covers most of the transmission distance. It can reduce the cost greatly and saves the trunk fibers by over 90%. 

 

It is also important to introduce mechanisms to flatten the network, especially at the network side, to satisfy the latency requirement of 5G bearing. Usually, the OLTs in the AO have limited forwarding capabilities, so two-layer forwarding exists on the OLT layer and the BNG layer. In the future, a more powerful OLT will bring the AO stronger capability to merge the BNG forwarding plane, while leaving only the control plane on the upper layer. In such way, the two-layer forwarding could be simplified to only one-layer. What’s more, the transmission functions could also be merged into AO so that only “one hop” is needed between AO and DC. By doing this, the processing time could be greatly cut down since the traditional way requires multiple hops between the IP layer and the transmission layer, and each hop brings complexity and delay. 

 

Cloudification is a clear trend in the development for the IT and telecommunications industry. The idea of SDN and NFV attracts many operators, and they have actively researched this idea and considered to migrate their networks in that way. 

 

The key part of SDN/NFV is the software components, which should be deployed above the cloud infrastructure. Typically, the cloud infrastructure is located in a Data Center (DC) which is at a higher level and far from the end users. For some latency-sensitive or experience-sensitive services like big video, the delay in such a long path leads to a bad user experience. Since the AO is the closest to end users, it is reasonable to place the cloud infrastructure to avoid the complex transmission or data processing between AO and DC, hence helping enhance user experience.

 

Limited space is one of the problems we may face when we try to deploy a cloud infrastructure in the AO. Most of the AOs have limited room, that is to say, they can only accommodate existing OLT equipment, but adding extra equipment would be quite difficult. Even if the AO has available shelves and racks, it would still be impossible to install universal servers in the existing OLT shelf since the depths of these two types of equipment are different. Besides that, the extra power supply and heat dissipation are also the factors we need to consider before the deployment.

 

The pluggable blades could work as a “Light Cloud” in the AO. Such blades could be inserted in the slots of the existing OLT to solve the problem of room shortage. As one of the line cards, the power supply and heat dissipation will not be the problem. The carrier-class reliability could even be an advantage over the traditional external servers. As the infrastructure oriented to SDN/NFV and Mobile Edge Computing (MEC), light cloud could be leased to third parties as open capabilities. And this might be a new business model in the future.

 

One of the promising applications based on light cloud is Access CDN. CDN is a commonly used technology to enable the rapid distribution for video services and has already been popular in the industry. Currently, most of the CDN servers are deployed in an upper level in the network, e.g. near CR or BRAS. When video traffic is relatively low, deploying CDN in such locations is a high cost performance choice. If the traffic keeps growing, the valuable uplink resources between BRAS and CR, and between OLT and BRAS would be inadequate and require upgrade. Deploying CDN to a lower level, like in the AO, would terminate large amounts of video traffic and relieve the emergence of uplink capacity shortage. Although deploying CDN on a lower layer may cost more for the CDN hardware, the cost reduction of the uplink resources may still make it a better choice. According to ZTE’s calculation, in a typical network model, deploying CDN in the AO would be wise in the aspect of finance, if the video traffic exceeds the certain level, for a two-level CDN system or three-level CDN system, respectively. Since it can be clearly forecast that the video traffic would raise dramatically, it is easy to expect the popularity of deploying CDN in the AO. At that time, the light cloud in the AO would be quite easy to deploy.

 

Another possible application is NFVI, such as vSTB, vRG, and vCPE. For example, STB can be decoupled to physical STB and vSTB. The physical STB provides basic function locally. The vSTB can be deployed in built-in blades to provide service application and dynamic resource allocation in cloud, shorter TTM and better user experience would be guaranteed since the location is close to end users.

 

In order to carter for the trend of AO re-architecture, ZTE developed TITAN as the integrated optical access node in the AO. TITAN is based on advanced full-distributed programmable architecture, supports end-to-end network slicing and the evolution to SDN/NFV, supports across-generation PON technologies from 10G-PON to 50G-PON, meet the requirements of both household broadband and 5G bearing. TITAN also innovatively provides built-in blade servers as NFVI to implement light cloud, such as built-in MEC, Access CDN and vSTB, and it will definitely build a convergent access office for operators.

 

In a word, the new development trend of the industry boosts the AO re-architecture. In order to operate new applications or meet the challenges of the new architecture, the AO in the future should include highly integrated PNFs and light cloud infrastructures. The highly integrated PNFs help the AO suit different scenarios including 5G fronthaul. While the light cloud infrastructure helps the AO better deliver experience-sensitive services including big video. ZTE TITAN helps operators build an agile AO for the future.