Network automation (specifically within the data center), is one of the most marketed, yet least realized of the product feature promises made by networking vendors. Networking vendors market automation due to the growing negativity associated with the manual configuration required of traditional networks, and the unique skill sets required to configure and operate data center networks. Enterprises require a more agile, simple to manage network to derive operational value and keep up with the changing requirements in the data center.
The data center world in which networks live has radically changed. The configure once, “set it and forget it”, reliance on the command line interface (CLI), the single group ownership, and the strict adherence to cable diagrams, are networking best practices where Enterprise applications changed every 12 to 24 months.
Unfortunately, even though networks are super-fast (evolving from 1Gb to 10Gb, and now increasing throughput to 100Gb and even 400Gb) and super-reliable, the network has become a boat anchor weighing down Enterprise IT. In modern data centers practicing agile development, application changes are happening all day long; traditional networks cannot accommodate this pace of change, and the networking team often is blamed for holding back the pace of innovation and deployment. To be clear this is not a network bandwidth issue, but rather a fundamental architectural mismatch in the era of orchestrators like OpenStack, VMware and increasingly containerized applications.
This is where the need for automation arises, and where the challenge of traditional networks are most evident, as at their core, networks were not designed for this pace of operational change. There are 100’s of configurations that are required in deploying, securing, and operating the network. And these are very dependent on the applications, tenants, service level agreements and communication policies within the application community. There are Day Zero cabling tasks, Day One network/system bring up tasks, Day Two daily runtime tasks, and Day N compliance and change management tasks. All of these can be automated, however there is no universal configuration that can be applied.
In order to achieve automation, modern data center networking solutions are deployed when installing a new rack with top of rack switches, such as a rack of hyper-converged nodes and the need to add capacity quickly based on unforeseen application demands. This new approach to networking automates Day Zero deployment requirements, including intelligence around physical cabling, physical port and resource type levels (server, storage array, leaf switch etc)
With next-generation networking, Day One automation delivers link aggregation, fabric formation and global configuration; instead of manual, switch by switch configuration via CLI, modern networking solutions manage these very important steps to get the network fabric operational quickly. Day Two automation is delivered via integration with the developer community where they can essentially configure a private network (in a similar approach and look & feel to an AWS VPC instance, delivered on-premise) through a self service portal. This is the area of automation most desired by Enterprise network operators, and often where the issues of traditional networking approaches are most visible and painful. With a next-generation networking solution, Day N operations are where the networking team can automate patch and security updates, automate compliance tweaks, and update external systems including CMDB’s etc.
As with any buzzword within this industry (remember Content Defined Networking, or Directory Defined Network, or now Intent and Cognitive Based Networking) the vendor community is liberally applying automation within their product collateral. As explained above automation is a broad name for many different networking tasks. Yes automation can make the network more responsive, whether it is for faster racking and stacking to address capacity issues, allowing the application community to become more self-sufficient as they embrace CI/CD, and/or handling compliance requirements based on security and regulatory enhancements that need to be rolled out quickly. The good news is that customers are beginning to acquire new networking solutions, while hiring and training their network staff, with a focus on network automation. Moreover, there are growing number of Network Automation Engineer job postings and there are now courses available, to enhance the CCIE core skill set, for example: https://www.ipspace.net/Main_Page
Recently, Big Switch was referenced in an article (alongside our partner Nutanix), Automation Is Tricky, So Get Started Now, that touched on what Big Switch refers to as Day One System Installation and System Bring Up Configurations. This automation for Nutanix Enterprise Clouds is significant -- it allows for the deployment of a hyperconverged rack of servers, cabled up to redundant white-box top of rack, multi-terabit switches, automated discovery and no touch configuration of the links between the switches and these servers. This includes both the physical links, as well as the virtual switch interfaces. Moreover, this includes more advanced features including link aggregation, link failover, and advanced telemetry. The Big Switch controller is reading in discovery data from the leaf nodes, and programming the links based on this data. It is relationship driven, based upon pre-defined relationship policies (those worked on with Nutanix).
The compelling part here is that customers can quickly deploy new racks of hyperconverged servers, and insure they are optimizing the network connections to their fullest extent without the need to have the network engineer on site. This frees the time of these networking experts to work on more important things such as capacity planning, multi-site cloud integration, or enhancing the security of the network (as examples).
The bottom line is that customers need to identify their network automation projects, in smaller digestible projects, with workflows that mirror the operations of specific functions, and that are possible without custom scripts, or rigid topologies.
Director, Product Marketing
Bill Erdman is a seasoned high technology product management & marketing executive with over 20 years of experience in data center communications. At Big Switch Bill manages product positioning, messaging and value propositions of software defined networking for data center physical (underlay) networks, multi-site cloud networking, and machine driven security and forensic analytics. You can connect with Bill on LinkedIn: https://www.linkedin.com/in/berdman/