How to integrate I/O virtualization into the network architecture

For IT administrators, there are many options to consider when setting network interfaces. It's easy to get started-a physical server with two network interface cards available ). But what should I do now? Based on the Environment size, engineers must also consider the application load, switching capability, virtual LAN VLANs), work traffic, and overall user traffic. In this article, we will discuss how I/O virtualization is integrated into the network architecture to help reduce hardware costs.

In the virtual network infrastructure, I/O virtualization solves problems related to network and interface capabilities. I/O virtualization uses a single high-speed I/olink connection for each physical machine, rather than having multiple interfaces and connections on each server as traditional practice. Its beauty is that the high-speed I/ochain is logically managed as multiple virtual resources. This is similar to running multiple virtual machines (VMs) on a single physical host. Virtual I/O enables administrators to generate multiple virtual network interface cards for network connection) and virtual adapter (HBAs) (used for Fiber Channel storage ). These virtual I/O cards can run exactly as accurately as the physical Ethernet and fiber-channel hardware they replace. The key is that because these network interface cards and virtual adapter cards are logically different, their network and storage connections remain logically different.

I/O Virtualization in a data center has the following benefits:

• Reduces the number of I/O cards per server: virtualization eliminates the physical layer. Unlike physical routers and fiber-channel HBAs, virtual routers and HBAs can be dynamically generated and presented to VMs without the need to restart the underlying server. This means that the normal running time of the physical host is increased.

• Reduced connection lines: As IT engineers, we have to deal with connection lines in any case. Whether there is a direct need to track cable faults to components or replace the secondary links of remote servers, dealing with connection lines is a daunting task. Therefore, why not merge some functional network tasks? Combining storage and network traffic can improve the utilization of a given link, reduce costs, and simplify the infrastructure. Since each physical I/O can support and process the traffic delivered by the server theoretically, multiple connections are no longer needed.

• Improving the economic efficiency of data centers: hardware equipment requires capital investment. Connecting cards, connection lines, and server Peripherals can quickly consume budget. By using virtual I/O, the required ingress and HBAs can be deployed in smaller server software packages, and the data center can save space, cost, and power. Blade systems can also benefit from unrestricted connectivity, because this connectivity effectively removes the limit on the number of ports in some systems.

• Simpler interface management: graphical user interfaces of the virtualization platform, such as those in XenServer, vSphere, and Hyper-V, have now been developed to more precisely manage the network architecture of an environment. For Junior and intermediate engineers, it is an important step to view the chart user interface. It can help them understand the virtual renewal situation, so that engineers can perform on-site changes without compromising normal operation, you do not need to worry about unplugging the physical connection.

As virtualization will continue to expand its impact in the data center, it is a continuous effort to find ways to improve network I/O. With the development of network I/O technology, the use of airborne virtual I/O technology will help meet the development needs of virtual infrastructure.

　I/O virtualization tools, tips and best practices

As mentioned above, each environment is unique, so they have their own set of requirements for network interface capabilities. Network Architecture is constantly evolving to meet new business needs, technological changes, and expected growth of end users.

To better understand the environment requirements, consider the following steps during I/O virtualization planning:

1. collection performance standards: performance data is an important factor when identifying which hardware is suitable for virtualization software or hardware. Administrators collect performance standards from servers, storage systems, and network devices, and the duration is no less than 30 days. Monitor the application and server running status through standard business processes, and then determine the peak and low usage of the application, as well as the type and quantity of resources used by the application. Remember, if your Server OS's built-in monitoring tool recognizes virtualization, you can also monitor it or seek professional consultants.

2. Use tools to help collect data: some software packages can be used to collect standard network data, which is easier to use than built-in operating system tools. Such products as Orion NetFlow Traffic AnalyzerNTA) enable administrators to accurately quantify how the network is used, Who is used, and for what purpose. NTA's plotting function associates traffic destined for a specific port, source address, target address, and Protocol with an application name that you can easily identify.

3. Understand network architecture and interface interconnectivity: A company often buys expensive network switches, it is found that the new machine does not bring any benefits to its existing server or connection environment. With a clear understanding of the actual environment, the administrator can understand what to buy. Generally, a medium-sized architecture can run well in a gigabit network. If you upgrade its internal components, the performance will not be improved. Network Research can show the resources used for internal bandwidth and whether problems have occurred.

4. ROI): Even if you migrate to a 10GbE environment or replace the NIC on the server, you must make sure that this investment is worthwhile. For mature IT managers, IT is a waste of money spending that exceeds the benefits of the tangible infrastructure. Sometimes, it is not worthwhile to do so, so before spending money, first understand your actual environment needs.

Technology is a constantly changing entity. The success of the technology depends on the end users who use the technology. A good IT manager is always happy to use existing and new technologies in a network architecture to adapt to the actual environment in which their needs continue to grow.