Internet Network Architecture Shows Cloud-Network Convergence Trend
The Internet has entered an era of massive traffic, massive connectivity, and wide interconnection. Simple stacking cannot solve the problems and challenges facing networks. Business demands and technological innovation jointly drive profound changes in network architecture. Inter-network interconnection, inter-cloud interconnection, and basic network evolution are the three main directions of architectural change.
The trend of efficient inter-network interconnection is evident, meeting the demand for improved quality and efficiency of Internet services
Global network connectivity continues to grow rapidly, and interconnection paths are becoming increasingly diversified and efficient. For a long time, the international inter-network architecture has been characterized by carrier-centric, strictly layered features, with Tier 1 carriers holding the discourse power in global network interconnection. With the vigorous development of Internet services, especially the widespread application of performance-sensitive services such as video and cloud computing, the problems of extra traffic detours and significantly increased costs caused by the layered architecture have become increasingly prominent. As a result, lower-tier networks have begun to seek extensive peering interconnection, and the strictly layered architecture has been broken. According to PCH's "2016 Global Peering Survey," 99.98% of interconnection agreements are peering between the two parties, with only 0.02% involving purchased transit services. Meanwhile, the proportion of low-connectivity networks is decreasing significantly. From 2011 to 2016, the share of low-connectivity networks dropped from 62% to 35%, a reduction of nearly half. In addition, the number of peering partners for the world's top 12 networks has increased notably, rising from 700–2,400 to 2,200–4,500 during 2011–2016.
As convenient interconnection channels, exchange points are expanding rapidly, accelerating the process of efficient inter-network interconnection. Exchange points are expanding rapidly worldwide, and the inter-network interconnection architecture is becoming increasingly flat. In recent years, global exchange points and their connected members have both expanded at a rate of approximately 20%, with more than 800 exchange points now operating across 147 countries. At the same time, exchange points attract a large number of network connections by virtue of their "low cost, wide coverage" advantages, with connected entities covering telecom carriers, Internet content providers, data centers, content delivery networks, cloud service providers, and domain/Internet monitoring infrastructure. PCH monitoring data shows that a large number of networks aggregate at exchange points. For example, the Ponto de Troca de Tráfego Metro So Paulo exchange point in Brazil has more than 1,200 connected networks; multiple exchange points carry traffic above 1 Tbps, such as DE-CIX in Germany with traffic reaching 5.2 Tbps. In summary, the network aggregation effect of global exchange points is significant. With the continued expansion of exchange points and the growing number of connected networks, the network architecture will be further driven toward flattening in the future.
In the multi-cloud era, interconnection entities extend to the cloud, driving the rapid rise of inter-cloud interconnection
Demand for network deployment targeting cloud interconnection is strong. With the maturity of the cloud computing industry, enterprises can choose cloud products from different companies according to their own needs and deploy multi-cloud strategies. According to the RightScale 2017 survey report, 95% of global enterprises use cloud services, of which 67% adopt hybrid cloud architectures; more than 20% of cloud enterprises connect to multiple public clouds, with an average of 1.8 public clouds per enterprise. In practical applications, the demand scenarios for inter-cloud interconnection are rich and strong, such as cross-account peering within a public cloud, dynamic adjustment of hybrid cloud services and computing resources, and network interconnection and service interworking between public clouds.
— Interconnection within a public cloud. Data exchange between tenants within a public cloud. For example, the finance department and accounting department of the same company each establish independent virtual private clouds with a cloud provider, but the two departments can access each other across accounts.
— Hybrid cloud interconnection. Some enterprises build hybrid cloud architectures, achieving optimal resource allocation at the enterprise level through division of labor between private clouds or self-built data centers and public clouds. For example, banks, financial institutions, and insurance companies must comply with regulatory requirements, completing transactions and transaction records in compliant self-built facilities while deploying applications and front-end systems in public clouds. Hybrid cloud interconnection requires the configuration of efficient interconnection paths to achieve efficient integration between public clouds and private clouds/data centers. Recently, hybrid cloud interconnection has become an industry hotspot, with some scenarios integrating with SD-WAN.
— Public cloud interconnection. Considering that cloud services are not absolutely secure and reliable, and to avoid the risk of being locked into a single cloud provider, many enterprises choose to deploy services across multiple public clouds, creating a demand for inter-cloud traffic exchange. However, public clouds are incompatible and do not interconnect with each other, and current multi-cloud technologies are relatively complex, making management, operations, maintenance, and upgrades difficult. Therefore, public cloud interconnection overall still needs improvement.
Cloud interconnection has attracted widespread attention from the industry, with the model led by cloud providers and realized through third-party centralized exchange gaining development space. Internationally, mainstream cloud providers have launched dedicated cloud interconnection platforms, introduced partner programs, and established extensive cooperation with network operators, exchange points, and transmission network operators. On the other hand, cloud exchange platforms have become important hubs for cloud interconnection, with cloud providers and enterprise data centers accelerating their interconnection with cloud exchange centers. In China, as domestic enterprises accelerate their migration to the cloud, cloud interconnection has also attracted attention from multiple parties, with development models similar to the international approach. For example, Alibaba has launched high-speed channels and established cooperative relationships with many domestic enterprises with network capabilities. In addition, traditional exchange points, basic telecom carriers, data centers, and cloud platforms regard cloud interconnection as a new business growth point, successively launching cloud interconnection products, and a number of cloud interconnection platforms have emerged.
Carrier network architecture optimization and reconstruction advance in parallel
Carriers continue to promote the optimization and adjustment of existing network architectures to adapt to the demands of key services such as video on networks, while reducing CAPEX and OPEX. First, to adapt to changes in content source distribution and network traffic models, carriers continue to push backbone networks toward flattening, with a large number of inter-provincial direct links emerging, significantly increasing interconnection directions between backbone nodes and blurring hierarchical boundaries. Second, with the development of video, metro network traffic has grown 5–10 times, requiring metro networks to be built with low convergence ratios, while multi-level architectures evolve toward three-tier architectures, eliminating aggregation switches and achieving metro network flattening.
Carriers are accelerating network reconstruction with the goals of DC-centricity, virtualization, and intelligence to enhance their flexible configuration capabilities and accelerate service deployment. In 2015, AT&T began deploying network function virtualization. In the past two years, domestic carriers have also begun field trials and commercial pilot deployments of vBRAS, vEPC, and vIMS. At the same time, carriers are paying close attention to network resource orchestration and increasing investment. AT&T is promoting innovation and cooperation in the management and orchestration field, driving the development of the ONAP open-source ecosystem, and China Mobile and China Unicom have both joined ONAP.
In summary, the development of Internet network architecture presents a cloud-network convergence trend, with obvious technology integration and multi-party cooperation. Building wide interconnection channels to efficiently connect content and users, networks at all layers also demonstrate the characteristics of flattening, virtualization, and intelligence.
Source: Internet article