Simplification Will Accelerate Software-Defined Data Center Adoption
by Steven Shalita
Organizations are looking for ways to scale-out their data center network to meet the increasing demands of the business. The digital enterprise, fueled by new applications, hyper-growth of the Cloud, and adoption of hybrid IT are all driving dramatic change. In order for organizations to fully realize the potential of technology, they are looking to modernize their network and leverage the Cloud for increased agility, speed and consistency. This means new approaches and capabilities are required. Ideally, all aspects of a data center would be virtualized and available as a service. We’ve already seen this happen with the compute and storage layers. But this has not really happened for the network.
The allure of the public cloud is the efficiency, elasticity and simplicity that the Cloud enables. It’s more flexible than the typical enterprise data center, and resources are readily available as-a-service. The harsh reality is that the “Cloud” is actually not less expensive than doing it yourself. In the end, the Cloud usually ends up costing more. In many cases, a lot more. What the business really wants is to have the agility and elasticity that the Cloud affords, while still being able to retain control and manage the enterprise as a unified environment.
Imagine being able to bring the simplicity, adaptability, and elasticity of the Cloud to the enterprise data center. This has been the promise of Software-Defined Networks (SDN). The Software-Defined Data Center (SDDC) allows bringing the Cloud-like scale and operating model into the data center to create your own private Cloud. Sounds great, right? But, why has this not been so easy to achieve? Well, it’s complex.
Yes, SDN has been all the rage. It promises to bring new flexibility and much desired change to the network. Disaggregation meant opening up the hardware layer and providing the IT organization more choice. After all, years of legacy network architectures were hardware bound and tied to software from a single vendor. Abstraction enabled configuration and operational flexibility. Automation centralized management to make it easier to provision and manage large-scale networks from a single vantage point. Programmability brought the benefits of orchestration for network capabilities and integrating control across platforms. With SDN, it would seem we were well on our way to the utopia of the next generation network.
But the true benefits of SDN have remained elusive. While SDN is all about change, and disruption to the status quo, it has come at a price. While many characteristics of SDN are net positive, one of the significant inhibitors to SDN adoption has been architectural complexity. To enable the desired common control, programmability and automation, many SDN solutions require centralized controllers and new protocols. The problem is that controllers increase complexity. They require wholesale changes to the existing network and limit the ability to deploy a single network across multiple locations. Controllers also increase latency, and inhibit network resilience. Fail-over times are slower because the controller, which works like an old-fashioned “supervisor” in a chassis switch, has to instruct connected switches what to do, and manages any and all cross-switch communications, provisioning and reconvergence.
To achieve wide-spread adoption, SDN must change. The controller needs to be engineered out of the network. There needs to be a drive towards simplification to make the network easier to implement and operate. There needs to be better interoperability with existing architectures to enable organizations to leverage their existing investments and drive a more graceful migration to an enterprise-wide SDN architecture without sending everything to the junk pile.
The network also needs to become more intelligent. The network matters – it sees all and knows everything. Therefore, the network can provide dynamic insights into how the enterprise is running. This will enable knowledge of application, network and end-point service state to better understand how the users and services are consuming the infrastructure, and conversely how the infrastructure is supporting the users and services. This will enable the network to dynamically compare actual versus desired state and automate corrective actions such as security or traffic policy changes, reroute traffic, and link to other systems to implement dynamic changes to the infrastructure, redefining real-time service assurance. We are almost there.
Finally, the simplification of SDN will enable organizations to bridge the operational models of DevOps and NetOps to overcome the growing skill gaps. It will reduce architectural complexities, enable new services, and allow changes to be implemented faster, and with greater confidence against configuration errors. The goal should be to dramatically simplify the operational model while enabling consistency across the network without compromise.
Steven Shalita has more than 20 years of industry and technology experience across enterprise and service provider markets with a strong background in service delivery architecture, service assurance, performance management, security, enterprise networking, MPLS, SDN and IP transformation projects.