Server virtualization has created software-defined environments for virtual machines (VMs), and the edge or last hop of the network now lives inside the server stack, providing connections between VMs and outside servers. Meanwhile, new programmable storage solutions can run over converged data center networks and must be correctly configured, provisioned and managed. Overall, the network infrastructure has become more programmable and it is quickly evolving.
Initially, when we discussed SDN, we referred to OpenFlow-enabled switches, but with the impact of server virtualization on the data center, virtual networks have become more popular. New virtual network solutions include the use of software overlay networks (OVNs) that rely on tunneling protocols, such as Virtual Extensible LAN (VXLAN) and Network Virtualization using Generic Routing Encapsulation (NVGRE).
A number of companies have allocated teams to research SDDC considerations and benefits. The following are five important aspects of SDDC to consider before investing.
1. Power is king: The metrics used to measure the success of SDCC are starting to look a lot like the metrics of cloud providers, where more power equals more dollars, and the cost savings on power will start to trump traditional vendor selection criteria. According to one Emulex customer, "The new metric is performance/cost/power," and this is changing how IT teams are evaluating IT solutions. As x86 processors become more energy-efficient, the benefits of hardware offload materialize in significant savings in power, heating and cooling.
2. Significant reductions in management complexity: When implemented correctly, SDN helps to reduce the management complexity associated with traditional data centers. For example, if a new VM or physical server is installed, all of the networking components around the new workload need to be updated. Traditionally, the next step is for the server administrator to open a work ticket to update the network configuration to support the new workload. This creates costs that may surface related to manual firewall reconfigurations, manual router updates, manual ACL updates throughout the network, etc. Server-based OVN SDN solutions, such as VXLAN or NVGRE, provide a way to automate network provisioning for migrated VM and workloads on bare metal servers. As noted above, OVN can significantly reduce network management costs.
3. Software stability: The industry is amazed at the ability of the Web giants to use free, open source software. But these supersized companies have a large staff of developers who are able to support open source software. Recently, the Open Compute Project (the community that is publishing open source hardware specifications) announced progress in providing an open source switch solution. These solutions work well for Web giants because they have the engineering-power to program switches, build control planes and rewrite software for specific needs. But for the enterprises that don't have the resources or the large teams of software developers, enterprise installations need to be built on proven and vendor-supported solutions while they keep their eye on open source solutions that may be more viable as they mature.
4. If you can't measure it, you can't manage it: CIOs should ask themselves three crucial questions: "How does SDN fit into my network infrastructure," "Are there any network diagnostics available to me" and "How do I span virtual and physical networks for diagnostics?" There are two things to note that are of importance here. First, providing a "smart" deployment that focuses on likely trouble spots will ensure that you have the data you need to troubleshoot critical SDN and application issues through deep diagnostics. An interesting use case for physical network SDNs is for integration of a monitoring network that is connected to key switches in the infrastructure. SDN is then used to direct flows to data capture devices at the click of a button, which reduces time-to-resolution for network issues. Second, lightweight diagnostics can provide an "indicator light" so you can be alerted to where trouble may be brewing. I have even had some companies want to send all of the lightweight diagnostics to a big data solution and start running analytics on it (e.g., Hadoop database).
5. Integrate management between OVNs and the physical networks: Anytime you virtualize, it is not easy to triage the source of problems, and as a result, performance issues can surface that may be difficult to find. You could have an over-congested link on a physical switch, but customers may be calling to tell you that their VM isn't performing properly. Integrated network management tools are essential so that it isn't hard to diagnose the root cause for network problems.
Overall, according to a GigaOM research study, 15% of organizations have deployed or are in pilot with SDN, and an IDC study forecasts the SDN market will reach more than $3.5B in revenues by 2016. In large part, this growth is because SDN is becoming essential to enterprise data centers, and in turn, CIOs will need to think about how this technology will impact data center design. From a strategic perspective, they need to think about how to build the right data center mix, considering servers, storage and networks in equal parts.
About the author:
Brandon Hoff, director of product management at Emulex, focuses on innovation and cloud technologies to enable customized solutions for cloud providers and innovative solutions for enterprise users. Hoff has focused his career on designing, developing and applying industry-shifting technologies to solving customer problems.
Is SDDC ready for mainstream?
There is a downside to SDDC
Software-defined infrastructure shapes compute resources
Dig deeper on Software-defined data center
Brandon Hoff asks:
Is SDDC the answer to a converged data center network?
0 ResponsesJoin the Discussion