What is a data center?

With the continuous development of industrial digital transformation, data has become a key factor of production, and the data center shoulders the heavy responsibility of data calculation, storage and forwarding, and is the most critical digital infrastructure in the new infrastructure. A modern data center mainly includes the following core components:

· Computing systems, including general-purpose computing modules for business deployment and high-performance computing modules that provide super computing power.

·Storage system, including mass storage module, data management engine, storage dedicated network, etc.

·Energy system, including power supply module, temperature control module, IT management module, etc.

·The data center network is responsible for connecting general-purpose computing, high-performance computing and storage modules in the data center, and all data interactions between them must be realized through the data center network.

Schematic diagram of the composition of the data center

Among them, the general-purpose computing module directly undertakes user services, and the physical basic unit it relies on is a large number of servers. If the server is the body of the data center, then the data center network is the soul of the data center.

Why is Data Center Interconnect needed?

At present, the construction of data centers of various organizations and enterprises has become common, but it is difficult for a single data center to meet the business needs of the new era, and there is an urgent need for the interconnection of multiple data centers. Mainly reflected in the following aspects.

Rapid growth of business scale At present, emerging businesses such as cloud computing and intelligence are developing rapidly, and the number of related applications is also increasing rapidly, and these applications are strongly dependent on the data center. Therefore, the business scale undertaken by the data center is growing rapidly, and the resources of a single data center will soon be insufficient. Limited by factors such as the land occupation and energy supply of data center construction, it is impossible for a single data center to expand indefinitely. When the business scale grows to a certain level, it is necessary to build multiple data centers in the same city or in different places. At this time, multiple data centers need to be interconnected to coordinate and cooperate to complete business support. In addition, in the context of economic digital transformation, in order to achieve common business success, companies in the same industry and in different industries need to share and cooperate at the data level, which also requires interconnection between data centers of different companies intercommunication.

Cross-regional user access is becoming more and more common. In recent years, the business of data centers has changed from centering on Web services to centering on cloud services and data services. The range of users of relevant organizations and enterprises has long been free from geographical restrictions. Especially when the mobile Internet is very popular, users expect to enjoy high-quality services anytime and anywhere. In order to meet the above demands and further improve user experience, qualified enterprises usually build multiple data centers in different regions, so as to facilitate the nearby access of cross-regional users. This requires that business deployment can span data centers and support the interconnection of multiple data centers.

Cross-regional user access

Rigid requirements for remote backup and disaster recovery Nowadays, people’s daily work is increasingly dependent on various application systems, and the continuity of these application systems depends on the stable operation of the data center system. At the same time, data security, business reliability, and continuity are getting more and more attention, and backup and disaster recovery have become rigid requirements. The real environment is an environment full of uncertainties and various risks. Data centers are always facing potential threats such as various natural disasters, man-made attacks, and accidents. By deploying multiple data centers in different places, business continuity and robustness can be improved. As well as the high reliability and availability of data, it has gradually become an effective solution generally recognized by the industry. To deploy backup and disaster recovery solutions between different data centers, the interconnection of data centers must first be completed.

Remote backup and disaster recovery

Trends in data center virtualization and resource pooling

With the gradual maturity of the cloud computing business model, all kinds of applications and traditional IT services are “going to the cloud”, and the cloud business is becoming a new value center. Therefore, the transition from traditional data centers to cloud-based data centers has become a mainstream trend. Virtualization and resource pooling are key features of cloud data centers. The core idea is to transform the smallest functional unit of a data center from a physical host into a VM (Virtual Machine, virtual machine). These VMs have nothing to do with the physical location, resource occupancy can be elastically scaled, and support free migration across servers and data centers, so as to realize resource integration within the data center and across data centers, form a unified resource pool, and greatly improve the flexibility of resource utilization. sex and efficiency. The interconnection between data centers is a prerequisite for VM cross-data center migration. Therefore, data center interconnection is also an important link to realize data center virtualization and resource pooling.

Virtualization and resource pooling

What are the options for Data Center Interconnect?

In order to better meet the needs of cloud-based data centers, many data center network solutions have emerged, such as Huawei data center switches (CloudEngine series), Huawei data center controllers (iMaster NCE-Fabric), intelligent network analysis platforms (iMaster NCE-FabricInsight), provides the following two recommended data center interconnection solutions.

End-to-end VXLAN solution

Data center interconnection based on end-to-end VXLAN tunnels means that the computing and networks of multiple data centers are unified resource pools, which are centrally managed by a set of cloud platforms and a set of iMaster NCE-Fabric. The end-to-end VXLAN domain of the user, the user’s VPC (Virtual Private Cloud, virtual private cloud) and subnet can be deployed across data centers, and can directly realize business interworking. The deployment architecture is shown in the following figure.

Schematic diagram of the architecture of the end-to-end VXLAN solution

In this solution, end-to-end VXLAN tunnels need to be established between multiple data centers. As shown in the figure below, firstly, the underlay routes between data centers are required to communicate with each other; secondly, at the overlay network level, EVPN must be deployed between leaf devices in two data centers. In this way, the leaf devices at both ends discover each other through the EVPN protocol, and transmit VXLAN encapsulation information to each other through the EVPN route, thereby triggering the establishment of an end-to-end VXLAN tunnel.

Schematic diagram of an end-to-end VXLAN tunnel

This solution is mainly used to match Muti-PoD scenarios. PoD (Point of Delivery) refers to a group of relatively independent physical resources. Multi-PoD refers to using a set of iMaster NCE-Fabric to manage multiple PoDs. Multiple PoDs form an end-to-end VXLAN domain. This scenario is applicable to the interconnection of multiple small-scale data centers that are close to each other in the same city.

Segment VXLAN solution

Data center interconnection based on Segment VXLAN tunnels means that in a multi-data center scenario, the computing and network of each data center are independent resource pools, which are independently managed by their respective cloud platforms and iMaster NCE-Fabric. The center is an independent VXLAN domain, and a DCI VXLAN domain needs to be established between data centers to achieve intercommunication. Moreover, the user’s VPC and subnet are deployed in their own data centers. Therefore, the business interworking between different data centers needs to be orchestrated by a higher-level cloud management platform. The deployment architecture is shown in the figure below.

Schematic diagram of the Segment VXLAN solution architecture

In this solution, VXLAN tunnels must be established inside and between data centers. As shown in the figure below, firstly, the underlay routing intercommunication between data centers is required; secondly, at the overlay network level, EVPN must be deployed between leaf devices in the data center and DCI gateways, and between DCI gateways in different data centers. In this way, related devices discover each other through the EVPN protocol, and transmit VXLAN encapsulation information to each other through the EVPN route, thereby triggering the establishment of a segment VXLAN tunnel.

Schematic diagram of a segment VXLAN tunnel

This solution is mainly used to match the Multi-Site scenario, which is applicable to the interconnection of multiple data centers located in different regions, or the interconnection of multiple data centers whose physical distance is too far to be managed by the same set of iMaster NCE-Fabric.

What key technologies are required for data center interconnection?

VXLAN is essentially a tunneling technology, which can superimpose a layer-2 virtual network on any routable network, and realize intercommunication within the VXLAN network through a VXLAN gateway. At the same time, it can also realize intercommunication with traditional non-VXLAN networks. VXLAN uses MAC in UDP encapsulation technology to extend the layer-2 network, encapsulates Ethernet packets on top of IP packets, and transmits them in the network through IP routing. Intermediate devices do not need to pay attention to the MAC address of VM; and IP routing networks have no network structure restrictions , with large-scale expansion capabilities, so that VM migration is not restricted by the network architecture. For more information about VXLAN technology, refer to “In-Depth: What is VxLAN?”. EVPN is a next-generation full-service VPN solution. EVPN unifies the control planes of various VPN services, uses the BGP extension protocol to transmit the reachability information of the second layer or the third layer, and realizes the separation of the forwarding plane and the control plane. With the in-depth development of data center networks, EVPN and VXLAN have gradually merged. VXLAN introduces the EVPN protocol as the control plane, which makes up for the lack of VXLAN having no control plane at the beginning; EVPN uses VXLAN as the public network tunnel, which makes EVPN more widely used in scenarios such as data center interconnection.