The switch is an important part of the data center network and is indispensable for the data center construction LAN. In the past few years, the data center has ushered in a period of rapid development. The network has become a part of restricting the development of the data center. It is necessary to transform the network, and the switch equipment has become the vanguard of the reform. The data center puts forward a lot of new requirements and functions for the switch. The original three-layer forwarding switch can’t meet it at all, but no one can give up this market. To meet the needs of different data centers, the switch has new features, which will switch. Several subcategories have been subdivided, and this article will detail these new data center switches.
vSwitch
The vSwitch (Virtual Switch) is a virtual switch that implements the network function of the physical switch. The vSwitch runs in the hypervisor of the virtual server. The traffic between the VM and the traffic between the VM and the physical network are forwarded through the vSwitch. The vSwitch forwarding behavior is complete. Controlled by the SDN Controller. The vSwitch implements traffic isolation and mutual access between different VMs. These accesses can be completed through the vSwitch inside the server. They do not need to be forwarded to the hardware switch through the NIC. Then when the VM is deployed on one server, the vSwitch is very There is a need to exist. vSwtich has flexible configuration and strong scalability. A common server can be configured with dozens or even hundreds of virtual switches, and the number of ports can be flexibly selected. For example, VMware’s ESX host, one server can simulate 248 virtual switches, and each switch has a preset virtual port reachable. 56. The vSwitch can be used to achieve performance by using a hardware switch, such as the Microsoft Hyper-V virtualization platform. The connection speed between the VM and the vSwitch can easily reach 10 Gbps. Of course, vSwitch will also bring new network problems, such as limited traffic visibility. Traffic between the same host VM will not flow out of the server into the physical network. New network policies need to be implemented, and I/O bandwidth pressure caused by VM migration. The vSwitch configuration needs to be manually repaired. The vSwitch also causes conflicts between virtualization management and network management. The vSwitch is inconvenient to manage. The administrator can only control the uplink port of the physical NIC on the host and cannot control many virtual ports on the vSwitch. These problems require new solutions. If the solution is not good, the vSwitch is difficult to develop. In any case, vSwitch has been used as a kind of switch in the actual data center, and has a certain scale of application, especially in networks with more VM deployments. The vSwitch has better support for VMs. Promote the application and supplement the gap in the application of the physical switch.
AISwitch
AISwitch is probably the hottest switch nowadays, namely artificial intelligence switches. AI has penetrated into every aspect of the data center, and the switch can’t be absent naturally. What new changes can the AISwitch cross-border combination bring to the network? The first is traffic analysis. In the past, traffic analysis was performed by mirroring the data on the switch to the traffic analysis device. It not only occupies the network bandwidth but also needs a separate traffic analysis device. If these can be done inside the switch, can you cut costs? To do this is also very simple, you can complete the installation of a GPU inside the switch, the GPU has powerful computing capabilities, integrated data analysis software on the GPU, you can directly complete the traffic analysis on the switch. Secondly, the switch has the computing power and can perform statistical analysis on various state data running on the switch to perform fault prediction and processing. The AI technology can be used to compare the running state of the device with the results of the massive data sample learning. The sample covers many fault types, and the AI learns to output the fault self-healing scheme based on these data. The CPU of the third switch is relatively weak, and the route calculation and various algorithm implementations cannot be too complicated. With the GPU and AI, the switch can make some complicated algorithms. For example, the switch HASH algorithm is basically based on the quintuple feature of the message. Many details of the packet are not able to participate in HASH. The simple purpose of the algorithm is to achieve fast forwarding. The longest route matching is the simplest HASH. How to find it. With AI, you can learn the algorithm based on the data forwarding situation on the switch, optimize the algorithm with low efficiency, and then solidify the algorithm into the hardware. In short, with AI, the switch is plugged into the smart wings, the switch is no longer a fool-like device for simply forwarding data, it also has its own brain, with highly intelligent equipment. AISwitch can realize a series of intelligent behaviors such as automatic forwarding, self-management, self-healing and fault handling. It is network equipment that operates automatically and without human intervention. It is foreseeable that AISwitch will have a very broad market development space in the future.
White-box switches
The white box is not a new concept and has been widely used in the server field. The white-box switch decouples the switch hardware and software. The user only purchases the switch hardware and then can use it with the software as needed. The biggest advantage of the white box is that it can reduce the networking cost. The technical concept of white-box switch coincides with SDN, and it has a natural fusion relationship with SDN. The emergence of SDN has once again brought white box switches. Because the operating system software is not installed on the white box switch, the price is much cheaper than the traditional switch, and most of the white box switches have a relatively high port density, which is very suitable for high-density network deployment of large enterprises, such as having large-scale data. The center’s Facebook, Google, etc. all use white-box switches to deploy the network and can customize white-box switches and select the appropriate operating system software according to specific network requirements. After the emergence of SDN, the demand for white-box switches is increasing. Because SDN requires a large number of white-box switches instead of traditional brand switches, SDN can complete the data center network through software controllers and white-box switches forwarded by direct flow tables. Deployment and network deployment is fast, low-cost, and easy to maintain, which is beyond the reach of existing data center networks. Both SDN and white box represent the evolution trend of the future network and are the cornerstone of future network construction.
These new devices have given network switches more functions, making the network more intelligent and cost-effective, and recruiting points to point to the painful problems faced by the real data center network. It is not enough for a traditional switch to only perform data forwarding. vSwitch stands for virtualization, AISwitch stands for intelligence, and White-box switches stand for low cost, separation of forwarding and control. These are the main directions for future network development. Each type of switch is a product of network technology development and their arrival. It is catering to the needs of today’s data center for rapid development.
