Responding to cloud business challenges: A new consideration of data Center network architecture

The achievements of modern science and technology make the human organs in space has been greatly extended, if the telephone is the extension of human voice, television is the extension of human vision, the data center is the extension of the human brain, then the network is a variety of extensions, as well as between the extension and the subject exchange information between the transmission of the instructions of the nervous system. The normal nervous system is one of the basic signs of a healthy human being, similarly, the large bandwidth, the low time delay, the high reliable network system is also makes the human various organs to extend effectively, enters the cloud age the basic guarantee.

Cloud business challenges to the network

As information services enter the cloud era, more and more users through a variety of different types of business equipment into the network, a variety of business equipment, the exchange of information between many users in the rapid growth of various types of access equipment and the number of devices, physical signs are also growing rapidly, how to these large number of The rapid and accurate transmission of different types of information among various types of service devices has become a huge challenge for the network.

The Stack service business processing method in traditional network architecture

As the terminal equipment providing service business directly, the designer has developed and used the corresponding physical interface and communication protocol according to its business characteristics. For example, Ge/10ge, FC, IB and many other physical interfaces and Ethernet, IPV4, IPV6, FCoE, Iboe and other communication protocols, each protocol has its own specific data format, and the corresponding control message. In the traditional network architecture, various devices in the network are forced to support these interfaces and protocols from both hardware and software in order to connect the terminals of different service businesses. At the data level of a network device, each received data must be identified by type, and its corresponding forwarding rules are found and forwarded according to the result. At the control level of the equipment, the protocol of each kind of business needs to be handled separately to coordinate and control the forwarding behavior on the data level. In this way, a number of overlapping business logic networks are formed in a physical network. As a result, as the number of service business types increases, the business logic network becomes more and more complex, and the software and hardware become more and more complicated, which leads to the decrease of forwarding efficiency and the development time of new business. In the data center, not only to support the tenant's business logic network, but also a single tenant may also exist in a number of independent departments of their own business logic network, resulting in a data center of the physical network needs to support a large number of business logic network, making traditional stack services difficult to burden.

New ideas of network architecture in data center

How can an actual physical network with a large number of nodes support a large number of business logic networks, and each business logic network may have its own address coding system and delivery rules? Many examples of life have told us the answers we need.

Postal code and standard mail encapsulation are widely used in the current postal system. Postal code is the address coding system within the postal system and the address of each business service point within the postal system. The postal code on each message, indicating the address of the target business service point. For the sorting and transferring in the postal system, simply identify the target zip code. Therefore, the postal code is to the user actual physical address encapsulation, realizes the user address and the transfer address the decoupling, solves the user physical address quantity rapid growth question. At the same time, non-standard postal items are standardized packaging, but also for the mail delivery process using machines to achieve automatic high-speed sorting is possible.

In the cargo transportation system, the standard container size and the standard target address have greatly simplified the intermediate link in the transportation process and improved the working efficiency.

As with the postal and freight transport system, the role of the network is to transfer the user's information from the source address to the destination address. The information of the address system and the format of information will vary depending on the business, but the information transfer function of the network will not change. An efficient network information transmission system will not and should not be affected by the transmission of information address system and format. The network information transmission system only in its marginal business node to be passed the information to do standardized processing, and the system internal transfer process is not affected by the content of the transmission of information.

From the network's overall architecture, the network information transmission system is like a bridge in all directions, connecting the network's Edge business nodes. The source node of the edge business needs only a variety of business requirements, add the target Edge Business node address, to the bridge to run the transport. And the bridge only needs to run one type of efficient transport vehicles, loading standard containers.

Bridging network separates business function from forwarding function, not only solves the influence of fast growing user address on network routing, but also simplifies the forwarding process between nodes in network and improves the efficiency of network.

Basic working principle of bridging Exchange network architecture

The Service access node of the network edge carries on the unified encapsulation to the different type data packets from the user, and adds the target and the source business node information, then sends to the node within the bridge network;

The internal nodes of bridge network are forwarded directly according to the address of the target business node, and the addresses used in the forwarding process are decoupled from the large number of actual user addresses.

The Edge target business node is packaged with a user packet received in a unified data format, and then the original data message is sent to the target user.

Compared with the traditional method of stack service business, bridging Exchange network architecture takes the result of routing protocol as the basis of data forwarding in the Exchange network, the service node address of the network edge is the address of data forwarding, the forwarding address of the data is decoupled from the user's real address, which eliminates the influence of the change of user address on the internal forwarding of the network. At the same time, the data transmission between bridge nodes uses only one standard format, which is advantageous to the high-speed and efficient transmission of data, realizes the protocol and function of lossless Ethernet such as DCB, and facilitates the simplified management of the network.

From the point of view of service business, bridging network provides point-to-point, point-to-point and multiple point-to-point connection services, the entire network is like a switch/router FC Exchange/ib switch? , the Business node port on the edge of the network is like the user port of the switch. For all data, simply configure the correct port address, and the data can be delivered over a bridging network without concern about how it is implemented within the network.

Selection of technologies used in bridging networks

The data center network already has a variety of network services and data forwarding separate solutions and corresponding protocols, such as GRE, Nvgre, VPLS, Vxlan, Macinip and other representative of the IT program, as well as TRILL,SPB, such as the representative of the CT program. For a data center network, in addition to the basic requirements of the network, it is hoped that the selected technology and standards can:

It is simple, convenient and dynamic to support point-to-point, point-to-point and multi point connection services.

Support the shortest path end-to-end;

As many logical networks and multicast and broadcast groups as possible to support the needs of many business logic networks;

Compatible with the existing data format of a variety of protocols to facilitate seamless interoperability of business;

It can not only provide CT solution directly, but also support various IT solutions effectively.

Typically, the data layer in the network determines what kind of work the network can accomplish under such a protocol, and the control layer enables the task to be done automatically and efficiently. The data layer plays a decisive role in selecting appropriate standard protocols.

From the data tier of protocols used in the current datacenter, IT solutions are based on existing two/three tier networks (such as the Ethernet multi-channel network), such as GRE, Nvgre, VPLS, Vxlan, and so on, and their solutions are designed to address the shortcomings and deficiencies of existing networks and for a specific problem. The essential problem of the existing two/three network is not solved effectively, and the potential resources of the network are not fully utilized.

The CT scheme, represented by Trill and SPB, tries to solve the basic problem of two-layer network in essence. Because in the concrete implementation, the CT plan needs to update the network equipment, this has the very big disparity with the fast Development data center market demand. Therefore, the CT solution must consider the problem of three-layer network, and how to support the existing IT program effectively, and work together with it in solving the basic problem of the two-layer network.

From the point of view of the supplier of CT equipment, the basic problem of two/three-layer network not only exists in the data center network, but also exists widely in the network system of Enterprise Campus network, operator's LAN, wireless access network and so on, it is of great significance to solve the basic problem of two/three-layer network fundamentally.

Bridging architecture constructs "resilient Cloud network"

Traditional network architecture, edge and core have similar functions; this kind of network architecture, if the network carries the cloud service, the server and the storage frequently changes, needs to the network whole network upgrades, in order to support the new characteristic, will cause the network maintenance complexity to be high; In addition, after stacking numerous complex businesses on the device, The performance of the network will also be greatly affected.

Drawing on the successful experience of the Internet architecture, Huawei puts forward the idea of "elastic Cloud Network", uses the Advanced Bridge-Exchange architecture, deploys complex functions to the edge of the network, maintains the simplicity of the core network, and adds the business control layer to strip the easily changing business parts from the network devices.

"Resilient Cloud Network" takes into account the high performance of the network and complex business processing, in the cloud brings a variety of new services, the whole network of equipment need not be changed, just upgrade the business control level, you can deploy a new data center business, network architecture long-term stability. For example: only to update the IPV6 control plane can be smooth migration to IPV6, to ensure the stability of the core layer, just upgrade the FC network control plane can do FCOE protocol upgrades, from fc-bb-5 to fc-bb-6, without changing the core layer of the network.

The influence of new technology on bridging network

Network switching is developing in the direction of optical transmission and optical switching. Because of the limitation of technology, it is difficult to modify the optical data directly in the light signal in a short time, so the direct transmission of the optical signal (i.e. End-to-end routing) is the simplest and practical technical scheme. The end-to-end forwarding mechanism in bridging network architecture can also adapt to the development of the future technology.

(Responsible editor: The good of the Legacy)