VMworld trumpeted VMware’s favorite topic, the software-defined data center (SDDC). Certainly VMware’s products are real and available today, and they are not the only vendor developing for SDDC. The question is: What is SDDC really? And Is SDDC an actual thing?
What is Software-defined Data Center?
The common definition for SDDC is something like “Software-defined data center (SDDC) refers to a data center where all infrastructure is virtualized and delivered as a service. Control of the data center is fully automated by software, meaning hardware configuration is maintained through intelligent software systems…”
Um, what? Let’s try that again.
The essence of SDDC is to build a server, networking and storage infrastructure whose resources IT can immediately adapt to different application workloads. Resources include server latency and IO setting, networking speeds and costs, storage priorities and performance, plus security and centralized management.
The base configuration is virtualized servers, networks and storage plus a business logic layer to communicate application requirements. The enabling communication layer is the application programming interface (API). IT groups applications into virtual resource groups that automate application security, performance, capacity and so on. Once IT has developed best policies, a new application will easily fit into the software-defined provisioning model.
Core elements of SDDC include:
- Server virtualization, which is already a mature technology.
- Software-defined networking (SDN), which is not as mature as its server counterpart, but which is making strides. Leaders in SDN development include the Open Daylight Platform and Neutron, the networking component of OpenStack.
- Software-defined storage (SDS), like SDN, includes virtualized storage but is not limited to it. In addition to virtual storage such as pooling, SDS includes automated provisioning and SLA tool sets.
On top of software-defined servers, storage and networks runs a layer of management and automation software. This is really what achieves SDDC, when IT can manage and control all three domains from a single interface.
- Efficiency. Virtualization extends efficiency up to a certain point. Virtualization can also be resource-intensive; one of the burdens in SDDC development is to lighten resource loads.
- Flexibility. Agile provisioning saves a lot of time and optimizes applications. It also makes it easier to fit applications to different venues like the cloud.
- Centralized control. Running the data center from a single console eases the burden on small teams, and lets large teams act in concert.
- Energy savings. Automating energy usage means that the data center will slow down or turn off unneeded resources, then spin them up as needed.
Is Software-defined Data Center for Real?
This all sounds good, and it is. But there is no hard and fast definition of what SDDC really is because there are so many possible implementations. This very quality leads to some question if SDDC even exists today.
Under our present technology, the only possibility of an actual SDDC is a single-vendor server, networking and storage infrastructure — and even then each of these domains must be fully automated and virtualized with APIs and master policies available to any enterprise application.
This is hard enough to do in homogenous settings, let alone multi-vendor ones. Standards would help of course, but are more likely to be found on the drawing board than in the data center.
Having said that, several vendors including Cisco, Microsoft and VMware are actively working on the SDDC concept. It is possible today to apply software-defined services to limited data center domains. For example, VMware works through its widely adopted virtualization model to enable unified platform products. These products treat the cloud (vCloud Air specifically) as an extension of the data center. In fact, the cloud is the defining architecture for their version of SDDC. VMware uses the cloud model for both the software-defined data center, which it treats as a private cloud, and vCloud Air. Their SDDC software logically defines the resulting infrastructure services and applies them to qualifying applications and services.
The users closest to achieving an SDDC model today are well-heeled, meta-scaled service providers like Amazon and Google, who already run highly homogenous environments. So even though widespread SDDC may not happen for a decade, we expect to see active product development and slow customer deployments that will accelerate over time.