As big data demands stretch established network infrastructures, IT architects have been adopting new strategies to expand the capabilities of enterprise networks. Big data requires rapid access to vast pools of data and the computing power to analyze that data, but conventional network design just doesn’t offer the bandwidth or the capacity. That’s why more big data infrastructures are increasingly relying on both server virtualization and software-defined networking (SDN). Although these two strategies have similarities, understanding the differences between them is important when designing an efficient big data infrastructure.
The value of SDN and virtualization has been documented. According to Infonetics Research, the global service provider market for SDN and network function virtualization (NFV) will grow from less than $500 million in sales to $11 billion by 2018, with NFV making up the majority of the market. IDC analysts also note that the growing adoption of SDN and virtualized servers will mean VARs need to rethink their sales strategies. Hardware will have a longer useful lifespan and become less mission-critical as the emphasis moves to software management. Those VARs specializing in selling services rather than boxes will be prepared for the coming boom.
Virtualization and SDN
So what are the differences between SDN and NFV, and what do those differences mean for big data deployment? Though both strategies impose software controls over the physical network, the differences are important.
Network virtualization enables network administrators to keep up with network changes by automating responses. Network virtualization provides an overlay that tunnels the connection between logical network segments. Rather than having to maintain a physical connection, software manages the flow service between virtual machines (VMs). That makes it easier for network managers to implement changes on top of the infrastructure without having to reconfigure specific servers or routers.
Once you have the capacity to connect virtual machines, you can use NFV to add functions such as firewalls and load balancing on top of the virtual connection. Just as administrators can create VMs in software, they also can create virtual functions such as firewalls or intrusion protection as part of the connection. Virtualization leverages x86 servers, which saves hardware costs and allows you to build an abstraction on top of the network. The result is fast, efficient service provisioning without manual configuration.
Software-defined networking goes further. While virtualized servers abstract services in the data center, SDN decouples the network control and forwarding functions from the underlying infrastructure, including applications and network services. SDN separates the control plane (which controls where the data goes) from the data plane (which forwards packets to their destination) to make it easier to dynamically provision an entire network using software. Data traffic relies on switches that are controlled using standard control protocol such as OpenFlow.
While virtualizations add functions to the physical network, SDN actually changes the physical network by providing a new way to provision and manage the infrastructure. Whereas virtualization can take place using existing server hardware and data traffic, SDN uses switches and requires a new network construct where the control and data planes are separate.
How Virtual Servers and SDN Contribute to Big Data
Both virtualized servers and SDN play distinct roles in big data systems.
Server virtualization makes it possible to scale services as needed to handle large volumes and various types of data for analysis. Virtualization has three specific characteristics that support scalability and efficiency:
- Partitioning – Multiple applications and operating systems can be supported in a single physical system by partitioning available resources.
- Isolation – Each VM is isolated from the physical system and other VMs, so if one VM crashes, the rest of the system isn’t affected.
- Encapsulation – A VM can be identified as a single file so it’s easy to identify based on the services it provides.
In the context of big data, what this all means is that the virtual system can scale to handle big data analysis, even if it’s unclear if there will be demand for extremely large data sets.
Whereas server virtualization makes it easier to handle big data analytics, SDN makes it possible to deliver data and computing resources from anywhere in the enterprise or in the cloud. SDN can be used to program the switches to deliver optimal data flow, thus ensuring better QoS between servers and cloud resources.
SDN makes it possible to manage traffic for large volumes of data. SDN is:
- Programmable – Network control can be programmed as a standalone function because it is decoupled from forwarding.
- Dynamic – By abstracting control and forwarding, adjusting for traffic flow across the network is easier for accommodating changing needs.
- Centrally managed – Network intelligence is localized in the SDN controllers, which maintain a global view of the network. Controllers appear to applications and policy engines as a single, logical switch.
- Open – Network managers can configure, manage, and optimize network resources quickly using automated SDN program. Because SDN is based on open standards such as OpenFlow, IT managers can write the programs themselves because they are not hardware-dependent.
So server virtualization and SDN both play a role in optimizing big data systems. Virtualized servers create more efficiency for data handling and analytics at the data center level, while SDN makes it possible to integrate an extensive set of computing and storage resources, including cloud systems, almost on demand. Taken together, virtualized servers and SDN deliver the agility and speed that big data systems need to be effective and deliver ROI.