A recent article by Lee Doyle in TechTarget includes a survey result of over 5,000 IT professionals who are asked about the biggest challenges in software-defined networking. The results (Figure 1) have some interesting implications for network automation and the visibility and analytics tools that support it.
Figure 1: Challenges for Building a Better Network (Source: TechTarget)
By far the plurality of answers is simply to make the network more agile, so that it can adapt to the ever-changing needs of the business. Second and third are supporting multiple tenants (which may be departments or customers) and managing the network centrally.
Only later do such considerations as faster configuration and faster provisioning come into play. Of course these are important, but it appears that they are mainly supporting functions to the larger goal of agility. This makes sense, as IT organizations of all stripes–compute, applications, security, and networking–take a more strategic approach to their mission.
Beyond Network Incrementalism
Does networking fall a little bit behind its IT counterparts in business relevance through agility? Gartner’s Andrew Lerner thinks so, noting that this is due in part to “network incrementalism” or “a strong preference for making small tactical, iterative changes with a focus on short-term benefits, over foundational and/or strategic changes that often provide longer-term benefits.”
It’s easy to see why this might be so. As noted in Gartner’s NetOps 2.0: Network Automation and Analytics to Stay Relevant in the Digital Business Era, “Networks are fragile designs, making it risky to introduce any changes when the results of a change can’t be predicted.” The demands on networks have led to increasing complexity, as they must be rapidly built to handle an ever-growing demand for many disparate connectivity devices and protocols at high traffic rates.
There are many ways that networks can be made more agile. Prominent academics such as George Varghese and Nick McKeown have called for formal design methods in the spirit of those that have increased the stability and agility of both software development and hardware design.
Questions from Everywhere
But it’s not as if networks are built and maintained in a vacuum. Network engineers, like all IT professionals, are beholden to business-wide needs and demands (Figure 2).
Figure 2: Business and Technical Requirements for Networks
To fully appreciate the problems that IT faces on an ongoing basis, imagine all the questions being asked by each department at once. Figure 2 also illustrates the reciprocal nature of business technology and culture–there are inextricable feedback loops between the two.
The value orientations of the business may drive the selection and development of technologies–an example here is the hunger for ever-better analytics and its role in using IT infrastructure as part of business planning. Similarly,technology may propagate changes in business culture: BYOD and the rise of devices needing to be authenticated and secured in the workplace is a common example.
Requirements from Above and Below
Networks, like all of IT, are beholden to business policies (intent) as well as to the state of the infrastructure (inventory, traffic, topology) at the time that changes are made. Both of these are important, as shown in Figure 3.
Figure 3: Supporting Business Intent While Remaining Aware of State
The functions of service orchestration and resource management, both of which support the applications that run the business, depend on:
Notice the other labels in Figure 3:
Visibility Supports Analytics
The data collected through the visibility infrastructure provides the basis for the analytics engines that help make the decisions for the constant loop of automation. These analytics engines, whether provided in a one-stop shop or partnered through an ecosystem, help ensure that automation fulfills its intent, and does no harm, by answering questions on the following topics:
An auto-remediation example with Splunk can be found here, and an application debugging examples using interface counters along with SLX Visibility Services is found here. We also have an application and infrastructure scaling example that includes Workflow Composer with the Docker Swarm API.
Similarly, a combination of automation and visibility can be used to mitigate attacks or to to load balance flows. Here, Flow Optimizer (which maintains a detailed traffic database correlated to policy) is used for both the visibility and the automation.
Flow Optimizer can also be used with both the SLX Insight Architecture and Workflow Composer. This combination provides closed-loop visibility to support a variety of sophisticated remediation use cases.
As the state of NetOps and DevOps matures, more and more tasks that could only have been achieved with desk analysis will be fully automated with formally verifiable, perhaps machine-learned, 100% closed-loop automation.
But as we move in that direction, the correctness of changes from the perspective of business policy will more than justify the operational costs of the network engineers who are business-driven and fully versed in the knowledge bases for network policies.
Contact your Extreme Representative for more information.