• 04/13/2015
    8:06 AM
  • Rating: 
    0 votes
    Vote up!
    Vote down!

How DevOps Helps Networks Scale

Enterprise networks are collapsing under increased application demands. DevOps principles and practices can lighten the load.

The load on IT is growing. Applications double every four years. Operational costs double every eight years. And data volumes double every 18 months. That's according to a March 2014 IDC Directions report. If you consider the increased rate at which disposable applications (like mobile apps supporting occasional events like the World Cup or the Final Four -- go Wisconsin!) are becoming the norm and the potential explosion of the number of "apps" on the network thanks to microservice architectures, those numbers might be considered conservative. Very conservative, in fact.

Either way, the reality is that the load on IT -- and increasingly on the network -- is growing. Quickly. That means that IT has to figure out how to shoulder that burden. In particular, the network has to figure out how to shoulder its share of the increasing burden, because by every account, the network is what's (still) in the way of achieving operational excellence. In 2014, EMA Research pinpointed "slow, manual processes to reconfigure infrastructure to accommodate change" as a key pain point for 39% of its respondents.

In addition, a recent Avaya survey focused on SDN expectations noted:

Eight-three percent say they have issues to do with service configuration, and this includes: dealing with the complexity of configuring services and applications; manual configuration of servers and switches; making virtual machine moves between data centers; poor integration of physical and virtual systems; requirement for maintenance windows; limited or complicated access control (including BYOD); and complexity of Spanning Tree and multiple network protocols.

Let's face it, scaling the modern data center -- and specifically its network -- is challenging. A large part of the reason for that is because we've spent years deploying point products to solve one problem or the other. The result is complex, fragile system topologies comprised of hundreds of individual boxes, each with its own unique set of CLI, GUI, API and management methods, each that must be balanced against unique upgrade, patch and maintenance schedules.

What a mess.

And now we're proposing even more variance and fragility by adding in applications based on microservice architectures and microsegmented networks and application services and -- well, it's no wonder that our backs are breaking under the weight.

Because we haven't learned to lift with our knees. The reality is that it's not the load that breaks you down; it's how you carry it. That's one of the things software-defined approaches to scaling operations associated with DevOps teach you: how to carry the load safely.

Sometimes that means sharing the burden by collaborating with them; the folks on the other side of the wall. Sometimes it means approaching the app deployment pipeline with an eye toward the processes that govern the pipeline flow and ferretting out inefficiencies along the way. And sometimes it means adopting automation to eliminate repetitious (and admit it, tedious) tasks that distribute the weight more evenly and allow you to take on a heavier burden.

A DevOps approach isn't just for developers or cloud, or for application and web server infrastructure; it's for network infrastructure, too. Whether that infrastructure is physical or virtual, there's a great deal of operational efficiency that can be gained from applying the principles associated with DevOps to the network.

We'll be exploring (and demonstrating) some of those principles in more detail at our 3-hour workshop Achieving Operational Excellence Through DevOps later this month at Interop Las Vegas. Join us on April 27 to learn more.


How to do DevOps

Lori, thanks for this post; it really puts things in perspective for networking pros, I think. I especially think the part about how you can apply DevOps is important -- through teamwork, ferreting out inefficient processes, or deploying automation. So much talk about DevOps refers to "culture," but I think in the network a lot of it revolves around those processes and automation. That does require a different mindset, but also great understanding of technology and how to operationalize it.

Re: How to do DevOps

@Susan, well said. The workload on the network is increasing on an exponential curve and at the same time, the world does not have the same level of individuals with the skills to meet the exponential growth, the only way to come out on top is to enable collaboration within a team -- gain the benefits of specialization and division of labor, etc.  


I think people hear "automation" and think "goodbye job." Is that a legitimate concern when it comes to the network?

Re: Jobs?

Great question, the words "automation" and "goodbye jobs" seems to be positively correlated in the minds of professionals. But the flip side is even more interesting i.e. as businesses in the market become efficient and the business next door has utilized automation to increase its efficiency then, wages begin to fall for the workers that are carrying out the same process in a manual fashion. And suddenly, "low paying jobs" are correlated with the question "why is this process not being automated?"

Re: Jobs?

So it sounds like you're saying that either way, folks doing manual but repetitive tasks lose? Do you see potential employment upsides to automation? I've heard the argument that automation means people can be freed up to do things that add more value or are more interesting. But I have to assume it also means you need fewer people.

And to be clear, I'm not advocating the preservation of manual tasks as a way to preserve jobs. I'm just trying to understand the full impact of automation in the data center.

Re: Jobs?

@Drew, the tradeoffs that you have mentioned are accurate. The potential employment upsides that can be gained through automations is similar to the upside that was gained during the first industrial revolution because, to a large extent the industrial revolution was due to the benefits that automation created.

The downside is that to labor is only valued if they have the specialized skills. If an individual has the specialized skills then, they will be able to do well in an automated industry but, if the skills are not as specialized as the industry requires then, it is bad news.

Re: How DevOps Helps Networks Scale

I tend to subscribe to the idea of DevOps being primarily a culture change. You need to look at it as an opportunity to cut out all the unnecessary processes that you've always wanted to get rid of, whether they're technology or people-based. You have to look at what your actual goal is, and tune every process to that. That's why it's such a great fit for software development, which has a clearly-defined and IT-related goal. Everyone can find some use there, though. Even if your organization is not going full DevOps, you could still take some cues from a "DevOps-like" approach.  The 'who should attend' field on your Interop presentation says that anyone building infrastructure to support next-gen networks and apps can gain value. Well, that should include just about everyone, but you're absolutely right! 

There are tons of great options out there for automation and central management to make this possible - hyperconverged appliances with automatic failover, open, low-cost NFV solutions that scale to meet demand, and more coming out all the time. In the same way that developers have to get used to eliminating or streamlining their usual test cycles, networking pros will have to accept some loss of granular control (actually, if you ask some vendors, you won't even have to do that) while not compromising on security or compliance. It doesn't have to be a bad or scary thing. If that's what your company wants you to do, doesn't it actually sound like a much less stressful way to work?