Five major issues of software-defined data centers

Five major issues of software-defined data centers

SDN has become the center of the "Software Defined Data Center (SDDC)" category, and it is mainly built on three important foundations: programmable, Software Defined server, network and storage.

Server Virtualization has already created a software-defined environment for virtual machines (VMS). Therefore, the last hop of the network boundary or network is within the Server Stack, it provides connections between virtual machines and external servers. In addition, new programmable storage solutions can run on an aggregation data center network, which must be correctly configured, allocated, and managed. In general, the network infrastructure becomes more and more programmable and is developing rapidly.

When we first mentioned SDN, we usually referred to a switch that supports OpenFlow. However, because server virtualization has a significant impact on the data center, virtual networks have become more popular. New virtual network solutions require software stack networks (OVN) that depend on channel protocols such as virtual extensible LAN (VXLAN) and NVGRE ).

Many companies have established teams to study SDDC issues and their advantages. The following are the five important issues they should consider before investing in SDDC:

1. Power supply is the top priority. The indicators used to define the success of SDDC become very similar to those of cloud providers, that is, the larger the power consumption, the higher the cost, and the power cost savings will become the main choice criteria of traditional suppliers. An Emulex customer pointed out that "new indicators are performance, cost, and power", which is changing the way IT teams evaluate IT solutions. With the increasing energy efficiency of x86 processors, the benefits of hardware reduction are mainly reflected in energy saving in power supply, heating and cooling.

2. greatly reduced management complexity. If implemented properly, SDN will help reduce the complexity of traditional data. For example, to install a new virtual machine or physical server, you must update all the network components required by the new workload. Traditionally, the next step is for the server administrator to submit a work order and update the network configurations used to support new workloads. The cost of this process may be far less than manual configuration of the firewall, manual update of the router, manual update of the entire network ACL and other work. The server-based ovn sdn solution provides a way to automatically allocate networks to migrated virtual machines and workloads on bare metal servers, such as VXLAN or NVGRE. As mentioned above, OVN can significantly reduce network management costs.

3. Software Stability: Some Web giants in the industry are incredibly capable of using free open-source software. But you need to know that these super companies have a large number of developers who support open source software. Recently, the Open Computing Project (community that released open source hardware specifications) announced the progress of an Open Source switch solution. These solutions are not often suitable for Web giants because they have enough engineering power to write vswitch programs, develop control panels, and rewrite software for special needs. However, for enterprises that do not have sufficient resources or large software development teams, enterprise applications must be built on reliable and vendor-supported solutions, at the same time, they are also paying close attention to open-source solutions. They may gradually become mature and become feasible solutions.

4. If the measurement fails, it cannot be managed.. The CIO should ask himself three questions: "How SDN integrates with my network infrastructure", "Is there any network diagnostic tools available now", and "how to span virtual networks and physical networks in diagnosis ". Pay attention to two important aspects: First, provide a "smart" deployment that focuses only on the most likely location where the problem occurs, this ensures that you obtain the data required to fix major SDN and application problems during deep diagnosis. An important use case of physical network SDN is to integrate a monitoring network, which connects important switches in the infrastructure. Click the button to send the flow to the data collection device, which reduces the time required to analyze network problems. Second, lightweight diagnostics can provide a "light", so you can receive alerts for potential problems. I even met such a company who hoped to send all the lightweight diagnostic information to a big data solution before running some analysis (such as Hadoop database ).

5. Integrated management between OVN and physical network: As long as the application is virtualized, we cannot easily track the root cause of the problem. Therefore, performance problems may be difficult to find the cause. There may be excessive congestion on a physical switch, but the customer will call you to tell you that their virtual machines are not running normally. The integration of network management tools becomes very important, which can reduce the difficulty of diagnosing the root cause of network problems.

In short, according to a research survey completed by GigaOM, 15% of organizations have deployed or are trying to use SDN, and an IDC study predicts that the SDN market revenue will exceed $2016 in 3.5 billion. To a large extent, this growth is because SDN is becoming an important part of an enterprise's data center. Therefore, CIOs need to consider how this technology will affect the design of the data center. From a strategic perspective, they need to consider how to build a correct combination of data centers, which should look at servers, storage and networks.