Zhang Haiping of Huawei: From the perspective of customers + industry, we will talk about the selection of core switches in data centers

Virtualization, cloud computing, and migration to colocation facilities are all having a significant impact on data center operations, and data center managers need to plan better technology solutions, consider security issues, and the potential impact of the Internet of Things as they respond to emerging technology trends. Data centers are facing a huge new transformation.

The development of data center networks and Ethernet standards has put forward higher requirements for a new generation of core switches, and many enterprises have an urgent demand for high-speed and intelligent data center core switches. So, how should enterprises choose data center core switches? With this question in mind, IT168 interviewed Mr. Zhang Haiping, Director of Data Center Network Marketing of Huawei Technologies Co., Ltd., and put forward suggestions for enterprise data center network construction and core switch selection based on his understanding of the trend of the times, Huawei’s strategic direction and innovative product solutions.

▲Zhang Haiping, Director of Data Center Network Marketing of Huawei Technologies Co., Ltd

Data center transformation in the era of big data

The core value of big data lies in mining valuable information from massive and complex data, and making faster analysis and more accurate prediction through big data technology, so as to discover new business models and create new business development opportunities. How to make data-driven business development is a problem that data centers must face in the era of big data.

At this stage, new technologies such as cloud computing, big data, Internet of Things, 5G and mobile Internet continue to promote the development of traditional data centers to high energy saving, high stability, high security, high scalability, high density, etc., which puts forward higher requirements for the scale and quality of data centers, and also brings greater challenges to the construction and operation process of data centers. In the era of big data, traditional data centers focus on responding to the needs of business departments, and their basic capabilities are no longer enough. The complexity of data means that data centers need to respond more quickly to changes and uncertainties in business needs.

In response to the above problems, Mr. Zhang Haiping said that the future data center network should be business-driven and provide the best IT services and experience for business departments.

He further elaborated that in the future, the core application of data will be to help enterprises discover greater business value. How to use big data, how to mine valuable information in massive data is the top priority, so enterprises should focus on the hidden value of data centers, fully tap the core value of data through integrated data platforms, continuously optimize data center business processes, reduce management costs, assist enterprises to make accurate scientific decisions supported by data, and contribute to the continuous innovation and development of enterprises.

The dramatic changes facing data centers have led to different diverse new needs for customers in different industries. As far as the industry is concerned, large Internet enterprises, telecom operators, financial institutions, large enterprises and government agencies are the main forces promoting the construction of domestic data centers, and they all have their own industry demands for data center network construction. For data center managers, they should actively embrace this change, while integrating the organizational structure of the two departments of network and IT, operation and maintenance personnel must improve from skills to quality in all aspects, in order to well meet the changes in business models brought by cloud computing.

Core exchange becomes the key Enterprise selection should adapt to new needs in the future

Mr. Zhang Haiping elaborated on the evolution process of data center switch architecture from the perspective of core switching architecture:

The first stage: bus-type switching architecture, switches based on bus structure are generally divided into two categories: shared bus and shared memory bus.

The second stage: CrossBar switching architecture, which can be divided into CrossBar+ shared memory architecture and CrossBar+ distributed architecture.

The third stage: CLOS switching architecture, CLOS is a multi-level switching architecture, in order to reduce the number of intersections in the middle as much as possible in the case of input and output growth, which is the mainstream architecture of the current data center core switch, is a new stage of data center switch development.

With the electrical signal switching rate reaching the bottleneck and the optical switching industry becoming more and more mature, the core switch switching architecture has entered the optical switching stage, and the current optical switching architecture is evolving from the internal switching of the data center core switch to the switching architecture between the entire data center internal switch, which is the latest practice of CLOS switching theory in the whole network architecture of the data center.

Under such conditions, the new generation of data center core switches should have some new characteristics: 1) CLOS switching architecture, cell exchange, VOQ queue; 2) Large-capacity elastic expansion, supporting interface types from gigabit to 10,000 Gigabit; 3) Support DCB, FCoE, FC and other technologies, so as to seamlessly integrate with the storage network; Support new interfaces such as CEE Enhanced Ethernet, so that Ethernet from the traditional “Best-effort” to a more mature “lossless network”, to achieve seamless integration with computing networks; 4) The programmability of the switch is extended from the control plane to the forwarding plane to quickly support the launch of innovative services; 5) Traditional services and SDN adaptation; 6) Switching, routing, transmission, switch color light and other multi-functions in one.

From the perspective of customer and industry development, he believes that the following five aspects should also be focused:

Business stability from the customer’s perspective: The cost of network switches accounts for about 10% of ICT investment, the cost difference per port between cell-based switches and packet-based switches is about 20%, and the switch difference accounts for 2% of ICT investment.

Smooth upgrade from the customer’s perspective: Smooth upgrade is a special scenario to ensure business stability, that is, under the requirements of localization or the introduction of new generation technology, smooth evolution is the core requirement. This is a robust option for customers without having to worry about whether the network will lose the quality of exchanges.

Network architecture from the perspective of industry chain: Data center network architecture has evolved from the past access and aggregation to the core to the two-layer architecture of Spine-leaf, which is recognized as the next generation of data center network architecture in the industry.


Digital transformation from the perspective of industry chain: Digital transformation is changing the business landscape at an unprecedented speed, and 64% of enterprises worldwide have become explorers and practitioners of digital transformation. Open networks have become one of the core powerhouses for enterprise digital transformation, and cloud-based data center networks are the core and cornerstone of network architecture in enterprise digital transformation.

Huawei’s star solutions help data center innovation

Mr. Zhang Haiping said that Huawei’s strategic direction for data center networks can be reflected in the following four areas:

Network pool: Based on SDN technology, network automation is realized, service deployment efficiency is greatly improved, and overlay solutions are provided in all scenarios, such as Network Overlay, Hybrid Overlay, and Host Overlay, which can meet the needs of various live network environments of customers, realize resource recycling, and network more flexible networking, greatly improving resource utilization. For example, if there are multiple server virtualization platforms on the customer’s existing network, the Network Overlay solution can be used to realize the heterogeneity of multiple server manufacturers and provide high-performance forwarding for various server access.

Security pooling: Supports standards-based service chains based on NSH, micro-segmentation based on more fine-grained security isolation such as discrete IP and virtual machine names, and makes it easier to automate security deployment, support distributed security, and support old traditional security devices.

O&M: O&M is not only reflected in the automation of deployment, but also includes the visibility, controllability, and management of network services, supports the mutual view of business, logic, and physical networks, and can quickly map path quality to the real physical network environment to realize the supervision of complex logical network environments. Real-time collection of traffic data of the entire network and monitoring of millisecond-level burst traffic packet loss can provide a basis for network active tuning. The intelligent fault location system based on the big data analysis engine, built-in anomaly detection algorithm and fault expert system database, can quickly locate the physical network fault node, and realize fault detection and rapid repair in minutes.

Openness: Based on the open standard architecture, it eliminates the barriers between cloud platforms, management tools, network devices, and network services of different vendors, and builds a full-level ecological alliance. Huawei has worked with more than 20 TOP partners and continues to grow to provide customers with more rich and complete data center network solutions.

Based on the above strategy, Huawei’s current representative solution in the data center is the CloudFabric cloud data center network solution, which aims to build simple, efficient, and open cloud data center networks for customers and support the long-term evolution of enterprise cloud services. CloudFabric solutions include automated deployment solutions, sophisticated O&M solutions, and open ecosystems to help customers generate real-time services, quickly locate and proactively repair faults, and quickly integrate application systems.

Based on Huawei’s flagship data center core switch CloudEngine 12800 series, high-performance box switches CloudEngine 8800/7800/6800/5800 series, and virtual switch CloudEngine 1800V, the CloudFabric solution works with Huawei’s agile controller Agile Controller to provide customers with a comprehensive cloud-based data center network using SDN technology. At the same time, it provides transmission, routing, security and network management product families to create one-stop data center network solutions for customers.

Huawei CloudEngine 12800 series switches are a new generation of high-performance core switches launched by Huawei for data centers and high-end campus networks. On the basis of providing stable, reliable, and secure high-performance L2/L3 layer switching services, it realizes elastic, virtual, agile and high-quality networks. CE12800 adopts Clos switching architecture, with a single slot of 11T, supports high-density 40GE/100GE line card line-speed forwarding, and supports 72-port 100GE on a single board, supports cell exchange, VoQ with 24GB ultra-large cache, and realizes non-blocking switching in a strict sense.