How the Cloud Data Center network architecture is merged

If only literally, data center Fusion network architecture is not so easy to understand, fusion itself is a hot word, many occasions are in use, from the three network fusion, technology integration to storage integration, everywhere in the fusion, then, what is the so-called fusion network?

Aside from the question of the definition of a converged network architecture, let's look at the development of data centers, especially the problems that cloud computing brings to the data center network.

Today's huge data centers abound, from Google to Facebook, where the size of the data center is staggering. At home, Baidu Data Center also has at least tens of thousands of servers, thousands of switches, and data center managers only dozens of people, the management of such a scale of equipment, network deployment automation requirements are relatively high. In addition, with the expansion of the business, there is also a high demand for network elasticity, which requires the seamless expansion of the network and the ability to perceive and support virtualization.

The traditional network architecture is almost powerless and its drawbacks appear gradually. From the technical point of view, the traditional network is mainly based on the Spanning Tree Protocol (STP), it was designed in the miniature hub era, even before the switch, the design is to ensure that a specific point of the connection has only one path, creating a loop-free architecture. Although almost all networks have redundant paths, all of these redundant paths are blocked through STP. As the environment expands, multiple routers and more paths are added to the schema, but STP still blocks all paths, leaving only one path.

When the active path fails, the network must be converged on the new path. In a large network, it takes a few seconds for the reintegration process to take. This can be accepted by standard IP traffic, but unacceptable for storage or converged networks, especially those with virtual environments. From the point of view of network bandwidth, STP is not extremely efficient. Because the blocked path means that the bandwidth is idle. In addition, active paths may not be the most efficient or shortest path between two devices. This will not only affect storage applications, but also the virtual environment, real-time virtual machine migration is detrimental, because the virtual machine or application, thermal migration (VMotion) may be required through several paths across multiple switches, and suboptimal path selection will only make performance worse.

In order to meet the needs of data center development, converged network architecture emerged. With the help of Cluster (switching), the VCS provides an integrated network architecture of distributed control plane with no host and no downtime, arbitrary connection and architecture Exchange.

In the VCS converged network architecture, the state, status, and configuration information (such as virtual machine (VM) metadata, network, and storage policies, etc.) between individual cluster switching nodes can be continuously synchronized, so it is easy to build an integrated network architecture based on self forming, automatic repair, and configuration. This greatly simplifies management and lowers costs and complexity.

By building the VCs virtual matrix, the cloud data center can easily achieve multi-tenant (user) service needs by effectively isolating and managing the data center's converged network infrastructure.

Now, with the maturity of product technology and the falling prices of products, it is time for data centers, especially large data centers, to bid farewell to the traditional network architecture, and an era of converged network architecture has arrived.