Three levels of energy savings in intelligent IP networks

The quest for better experiences continues to propel service diversification. For example, cloud VR lets gamers experience a new level of immersion, 5G UHD video gives surgeons the perspective they need for remote surgical guidance, and smart power grid O&M engineers use 5G private networks for video inspections and differential signal protection in distribution networks. The integrated application of technologies like 5G, AI, and IoT is shaping a world where all things can sense, all things are connected, and all things are intelligent.

With the 5G and cloud era fast approaching, we will see applications with tens of millions of users, imposing immense demands on network bandwidth and resulting in a predicted 10-fold jump in network traffic over the next 10 years. The resulting expansion in network capacity and increase in sites will drive growth in equipment energy consumption costs, which could counteract some of the benefits brought by new services.

The ICT industry currently accounts for a considerable amount of the world's total power consumption. For telecom operators, electricity bills typically account for a sizable portion of their total operating costs.

A large proportion of the vast amounts of energy consumed is unnecessarily wasted. The IP network converges mobile services, enterprise services, and home broadband traffic, and connects to data centers, acting as the basic bearer network. Given this, it makes sense to maximize network efficiency and build intelligent, energy-efficient IP networks.

The first step is to look at the IP network from a network-wide perspective and introduce intelligence with a focus on improving network-wide resource utilization. The next is to build an ultra-wide network to respond to the urgent need to increase service bandwidth, improve energy efficiency per gigabit, optimize network architecture and topology, simplify layers and network sites, and lower redundant consumption. Last, the concept of dynamic energy conservation should be applied to devices with components scaled out dynamically, enabling precise energy saving and cost reduction based on various application scenarios. Doing so can find the best balance between performance, function, and energy consumption.

Network-level energy saving

Intelligence can be incorporated in the IP network to optimize network traffic and resource utilization.

Traditional IP networks utilize shortest-path-algorithm routing protocols and best-effort data forwarding. This approach offers advantages in terms of accessibility, interoperability, and flexibility, but it can also easily lead to resource imbalances. There might be heavy loads on local links, such as backbone and metro ingress links, while the rest of the network links light loads. On the same network at the same time, loads may reach 80 percent or more on some links but only 10 percent on others. This can cause congestion and packet loss on certain links, which impacts service experience, while others experience low utilization and sit idle, wasting energy. Optimizing network-wide resources and improving utilization can significantly decrease power consumption and optimize the energy consumption ratio.

Optimizing network links and traffic first requires path adjustment capabilities to flexibly adjust paths based on various SLA requirements such as bandwidth and latency policies. But traditional traffic engineering, which involves manual planning and static configuration, is unsuited to handling complex traffic scenarios.

Based on different service requirements, Huawei's iMaster NCE + SRv6 enables intelligent routing, flexible and programmable network paths, and guaranteed connections. The solution also supports real-time visibility of network traffic status and automatic real-time adjustment of network traffic. An innovative ROAM algorithm provides capabilities of optimization based on multiple dimensions, such as bandwidth, latency, cost value, and priority, calculating optimal end-to-end paths for the whole network for optimal service paths and balanced global network traffic. Compared to legacy networks, the network utilization rate is increased by at least 20 percent and the overall network efficiency is greatly improved.

Site-level energy saving 

400GE builds ultra-broadband infrastructure networks and delivers optimum energy consumption per bit.

Traditional IP bearer networks are constructed with a focus on hardware. They’re generally divided into five or more layers: the access layer, the convergence layer, the metro layer, the backbone layer, and the service layer. Along with the continuous expansion in network scale, explosive growth in network traffic, and a substantial increase in the number of multi-layer devices, networks have become more complex and energy consumption has risen.

IP network interface rates have continued to increase in response to acute demand from ever-growing service traffic. At the access point, rates have risen from the gigabit level to 10GE/50GE, and metro and backbone rates have shot up from 100G to 400G. The next-generation high-speed interface technology, 400GE, uses 75 percent less optical fiber and consumes 20 percent less energy than 100GE, slashing transmission costs and power consumption per bit and eliminating load imbalances caused by link bundling on 100GE. IP networks with 400GE-ready convergence, metro, backbone, and data-center layers will help operators build ultra-broadband networks that provide an ultimate experience, and dramatically boost their return on investment.

With trends such as enterprise services moving en masse to cloud and mobile core network user planes and home broadband content moving closer to the user side, reasonable network planning should be DC-centric. By simplifying network hierarchy and the number of sites and devices, the energy consumption of the bearer network lowers proportionally.

In backbone sites, Huawei has integrated P and PE nodes using an integrated backbone solution, substantially cutting the cost of backbone network construction, reducing the network layer from two layers to one, and delivering overall energy savings of 10 percent.

In metro sites, Huawei uses metro fabric architecture to deconstruct the traditional metro router, separating network bearer and services. This allows the network to be flexibly expanded on-demand in scale-out mode, slashing the number of switched network components needed in traditional metro scale-up. Overall metro network construction costs are cut by 30 percent and energy consumption by 50 percent. The solution also provides large-capacity, non-blocking forwarding capabilities.

Device-level energy saving 

Serialized high-efficiency components + intelligent dynamic design enhances device energy efficiency.

The growth in traffic and pipe interface rates will inevitably lead to an increase in the capacity requirements of devices. With ever-larger router capacity, the power consumption of the whole device will rise significantly. Energy-intensive hardware will lead to a host of problems.

First, energy-intensive equipment will not only bring about higher power consumption and a sharp climb in operating costs, it will also produce a substantial amount of carbon emissions. Second, energy-intensive hardware imposes high requirements on equipment room power supply systems. Air conditioning and other support infrastructure also need to be upgraded. Third, excessive internal temperatures will impact the reliability and service life of energy-intensive devices. Statistics show that a 1-degree rise in ambient temperature increases component failure rate by 10 percent, greatly diminishing reliability and impacting the stable operations of equipment.

Huawei's NetEngine routers boast low-power components, efficient heat dissipation, and efficient power supply technology to break through limitations and decrease the overall power consumption of equipment.

Low-power chips 

Specialist experimental analysis reveals that 80 percent of the energy a router uses is for powering the line card, while 60 percent of the energy the line card consumes is used by the chipset. Therefore, device energy saving mainly depends on the power consumption of the chipset. NetEngine routers use low-power chipsets (under 0.4 W/Gbit), which consume 30 percent less overall energy than similar products, cutting carbon emissions by 30 percent over the industry average. A single router can save up to 180,000 kWh of electricity and 360 tons of carbon dioxide per year, the equivalent to 10,000 square meters of forest coverage.

Efficient heat dissipation 

Mixed flow fan + VC phase change heat dissipation solves air cooling limitations. 

Most mainstream devices adopt air-cooled heat dissipation systems. In most cases, a heat sink radiator and thermal pad are placed on the chipset and a fan remove the heat from the device, thus achieving heat dissipation. Therefore, the key components that determine the heat dissipation capacity of air-cooling systems are heat sinks and fans.

Copper heat sinks and silicone grease traditionally used to conduct heat have low overall thermal conductivity and average heat dissipation effects. 

Huawei does several things differently. First, we use a carbon nano thermal pad to convert irregular heat dissipation to directional heat dissipation, significantly improving thermal conductivity. Second, we use a vapor chamber (VC) liquid-gas phase change heat sink. The inside of the radiator has a vacuum chamber with a capillary structure filled with a refrigerant and a low boiling point, which quickly dissipates heat through the phase transition from liquid to gas.

The VC phase change heat dissipation and carbon fiber thermal pad technologies can increase chipset heat dissipation efficiency by up to four times and lower chipset temperature by 19 degrees compared to traditional heat dissipation methods, reducing heat build-up in the motherboard and greatly improving reliability.

One of the keys to determine the cooling effect of fans is air volume. The fan blades of typical fans cause airflow disturbances when inhaling air, which affects the amount of air inhaled. Huawei's mixed-flow fan uses a special fan blade design that decreases airflow disturbance and turbulence near the fan blades, tripling air volume. Fan efficiency is increased by over 10 percent with the same air volume, saving 200 to 300 W per fan tray assembly.

Moreover, adjusting the maximum power of traditional fans is difficult. When the overall power usage of equipment fans is high, power distribution requirements on equipment rooms are also high. Huawei's mixed-flow fans offer software-defined maximum power, flexibly adapting to fan power requirements, and reducing power distribution requirements on the equipment room.

Efficient power supply

There are three levels of conversion from the external power supply to the power supply unit of equipment motherboard components. Traditional first-level power supplies adopt N + N power backup, which not only takes up more space but also sets certain requirements on the external power supply. Huawei uses the dual-input power supply module with millisecond-level switching. Adopting N + M backup mode, it offers a substantially smaller power module and a 90-percent power efficiency improvement.

The module also uses magnetic blowout technology, enabling fast millisecond-level switching. Backup power supply switchover time is under 6 milliseconds, providing superior power supply reliability.

Multi-level conversion results in a loss of energy, so a 1-percent improvement in conversion efficiency in a single piece of ultra-high capacity hardware, which can exceed 10,000 W, can save nearly 1,000 kWh of electricity a year. Maximizing the power conversion efficiency of each piece of equipment can, therefore, have a big impact. Huawei's NetEngine 8000 router products support AC, DC and HVDC hybrid power supply modes and can increase the forwarding rate by 4 percent.

Intelligent dynamic energy saving

IP network traffic is bursty, with high loads when equipment is busy and low loads when it’s idle. This has led to dynamic energy-saving technology, which is used in Huawei's NetEngine routers. As well as offering the typical dynamic shutdown of unused line cards and ports, Huawei's NetEngine routers are also specially designed with dynamic energy-saving technology for traffic. This allows the number of working network processor cores and clock frequency of chipsets and the number of SerDes buses to be adjusted in real time according to traffic, maximizing the ability to lower energy consumption in various traffic scenarios.

Energy conservation is a gradual process that needs to be considered alongside other factors such as current network situation, evolution trends, and cost. To choose the right architecture and evolution strategy the whole network must be taken into account from a range of dimensions, including network, sites, and equipment.

For the full-service intelligent era of 5G and cloud, Huawei will pursue network construction concepts like intelligent super capacity, intelligent experience, and autonomous driving in its data communications products, and continue to innovate to help global operators build more energy-efficient intelligent IP networks.