ENTERPRISE MANAGEMENT ASSOCIATES®
New EMA Research Reveals Enterprise Strategies for Data Center Network Transformation
It's no secret that enterprise IT organizations need to change the way they do data center networking. Yesterday's networks were built to support static and monolithic applications. They were managed manually and optimized for north-south traffic patterns. Today's applications are dynamic and multi-tiered, and they span multiple data centers and clouds. They require agility, virtualization, automation, and any-to-any connectivity.
This summer, EMA launched a new research project titled, "Data Center Network Transformation." We surveyed 200 IT professionals directly involved in their organization's efforts to reinvent data center networking. The results will be published this fall. This column offers a preview of our findings.
Data Center Network Transformation Drivers
The technical initiatives driving data center network transformation suggest that enterprises are paving the way to digital transformation. Big data analytics, a central component of data transformation, emerged as the number one driver of data center transformation (42% of respondents). Hybrid cloud and multi-cloud architectures (32%) and the Internet of Things (31%) were also top drivers of change.
We were surprised to discover that DevOps (12%) was the least impactful drive of data center network transformation. Many of the transformational networking solutions on the market today tout their ecosystem integrations with DevOps tools and practices, suggesting the DevOps should be a major driver of change. On the other hand, further data in this report suggest that data center network transformation projects do tend to align with DevOps, even if DevOps isn't a major driver. For instance, 55 percent of research participants are using DevOps automation tools (e.g. Puppet, Ansible) to automate data center networks.
SDN is Essential to Network Transformation
Ninety-eight percent of organizations involved in data center network transformation are adopting software-defined networks (SDN), including 36 percent who have completed a deployed a production SDN network and 44 percent who are in the process of deploying one today.
According to these organizations, the two most important SDN use cases are network virtualization (46%) and disaster recovery (44%). This suggests that these organizations are building multi-tenant networks, perhaps to serve different application groups or business units. They are also using SDN to logically replicate their networks in case of a data center outage. Secondary SDN use cases include automated network provisioning (29%), application-specific optimization (26%), and microsegmentation (26%).
Security risk (30% of respondents) was identified as the top technical challenge to SDN adoption. Many respondents also cited ineffective quality of service enforcement (20%) and insufficient interoperability with legacy networks (20%) as technical challenges.
Disaggregated Network Switches Gain Traction
Forty-five percent (45%) of respondents have deployed disaggregated (e.g. white-box or bare-metal) switches in their data centers, and another 41 percent are considering them. We have no insight into the extent of such deployments, or whether they are pilot projects or broad deployments. However, this finding suggests that many enterprises are at least experimenting with disaggregated switch hardware and software.
The number one benefit that IT professionals see with disaggregated switching is the ability to integrate network operations with server operations (35%). For instance, some network operating system vendors offer hooks in their software that allow IT organizations to manage them with the tools they use to manage server infrastructure.
Risk is the biggest barrier to disaggregated switch adoption. Thirty-eight percent (38%) say their security and compliance teams object to disaggregation. Also, 34 percent are troubled by divided support models, with a lack of unified support from network software and hardware vendors.
Virtualizing Application Delivery Services
Finally, we found that seventy percent of these data center network transformation projects involve virtualization of application delivery controllers (ADCs) and/or load balancers. When these projects are complete, IT organizations expect about 58 percent (on average) of their application workloads to be served by virtual ADCs.
The biggest challenge to ADC virtualization is cloud or network orchestration incompatibility (42%). Meanwhile, nearly half (46%) of these organizations are orchestrating overall ADC functionality via cloud orchestration systems like OpenStack. Research participants also told us that scalability (34%) is the most essential characteristic of ADCs used in their transformation projects. Also, 27 percent want ADCs that are resource-efficient. In other words, they need narrow feature sets that are tuned to specific application requirements.
Much More Data in Full Report
Be sure to check out the full report when we publish it. It will include insight into the types of tools that enterprises use to design, build, automate, monitor, and troubleshoot their transformed networks. We will identify the organizational changes and training priorities engendered by these projects, and we'll identify some best practices based on the transformation success rates our research participants reported to us.