Offering highly available and scalable cloud services at a cheaper rate than their competition -- as well as the cost to the customer of an in-house deployment -- is no easy feat for providers.
VMware says the software-defined data center
What are the main differences between a traditional data center and software-defined data center from the perspective of a service provider?
Neela Jacques: Managing a data center is still incredibly expensive from a capital cost basis, but also [due to the] army of people performing the same repetitive [administrative] tasks over and over again, which is a little silly and doesn't make sense. Providers in the industry have realized that while we have done a great job of abstracting pooling and automating the delivery of compute and memory, we have really ignored the networking, storage, security and availability. When a user wants to deploy a new workload in an environment or scale an existing workload, networking, storage, security and availability are still highly reliant on physical devices.
A software-defined data center is an architectural approach aimed at decreasing complexity for the provider by automating a lot of tasks that are done manually in order for the provider to be much more efficient with the way they run their environment. To accomplish that automation, we need to abstract every resource -- like compute, storage, networking and security --at the level of the data center and pool it together into what VMware calls "logical data centers" -- which means automating as much as possible on the back end.
Why should providers move towards a software-defined data center architectural approach?
Jacques: Moving to a software-defined data center approach offers tremendous advantages for providers in terms of costs. Providers won't have to guess how much gear they will need, and overestimate to cover their bases. The approach is also beneficial from a management standpoint. A software-defined data center allows the system to intelligently make automation decisions, freeing [data center administrators] to focus on other things. The number of systems that an administrator can manage can increase dramatically, allowing the provider to support larger environments and their customers' growing needs.
Could moving toward a software-defined data center approach boost a provider's competitive edge?
Jacques: If you look at service providers right now, they are competing against two formidable groups -- users running infrastructure in their own environment and also with and Amazon EC2. Many [customers] are becoming more efficient at creating more virtualized, software-defined networking (SDN) type environments, while Amazon is becoming more pervasive. There is huge pressure to deliver more [services] and lower costs at the same time, so a user's interest in a service provider is in being able to manage [a user's] environment at a lower cost, while offering agility and being able to innovate more in [their] offerings -- not just being able to provide infrastructure anymore.
With a software-defined environment, [providers] will be able to provide faithful service and free their own employees from spending all their time deploying a workload. These same people will now be able to focus on delivering a higher service-level agreement (SLA). This approach can help providers lower their costs by 20% -- savings they can either pocket or use to keep them competitive with other service providers.
What are the components of a software-defined data center?
Jacques: There are two major component categories. [Providers] need a better kind of infrastructure, because pooling and automation begins as the infrastructure layer. They will also need the ability to abstract storage and networking resources, along with compute.
More on software-defined data center
Is the software-defined data center ready for the mainstream?
Software-defined storage: Is hardware obsolete?
The downsides of a software-defined infrastructure
At the infrastructure layer, we are seeing … virtualization platforms like VMware Vsphere evolving and continuing to get powerful to deal with a broader set of use cases. The next thing [providers] need is the ability to start virtualizing the network. Providers could use vCloud Networking and Security, a tool that is all about providing a distributed switch to create a full, virtual network. I think we are also going to see some really interesting innovations in the next few years on the storage and network availability side -- like having different kinds of options in terms of where data can be stored, not just traditional arrays, but also distributed storage systems and internal disk and virtual storage area network (SAN)-like constructs.
The second big category is management. As providers begin pooling everything to a greater level, it needs to be managed differently because there won't be one person that can know everything that is happening on the underlying hardware or virtual machines. Instead, providers need to have a system or tool that looks at what happens in the environment -- like typical performance and throughput of switches -- and can report any major changes that may need to be investigated.
Are there heavy up-front costs involved in moving toward a software-defined data center?
Jacques: Yes and no. Providers are still buying hardware, so the upfront costs of starting up a software-defined data center may be lower than creating one around hardware because the provider will be more efficient in what they are buying and building. For some, they are going to have to invest up front to make money in the back end, but I think it would certainly be cheaper to do the same things cloud providers were doing before SDN. Cost savings are being reinvested with virtualization, and so we are seeing providers be able to offer more value and more services to the end user.
How will a software-defined architectural approach to the data center benefit service providers' customers?
Jacques: Service providers will be significantly more responsive to their customer's needs. The modern provider today should immediately be able to provision infrastructure for a customer without racking and stacking new hardware. The provider can also use analytics-based techniques to see when something is going wrong in a user's environment. A good service provider doesn't just stand up infrastructure for [a user], but is able to actually come back and tell [their customer] that they have extra traffic, and offer services to accommodate this change immediately. In many cases, it won't just be cheaper, but the provider can offer higher value to the end customer.