Skip to main content
Jun 20 2018

Never send a human to do a machine’s job

It's Math-omatic! 

Maybe the film famous quote: "Never send a human to do a machines job" works in the Matrix, but is it true for all use cases all of the time? What's the break-even point between automation and manual operations?    

In any production process, there comes a decision point for introducing automation.  What's the right time to start manufacturing my gadget with a mold? When's the right time to introduce a robot into the assembly line? Should it be a unique automation system per operation? Which new skills does my work force need to acquire? Is all of this really worth the effort (time & money)?  

Surprisingly similar questions exist (or at least they should exist) in the domain of telco services delivery, where it can become quite difficult to estimate the real value such a system brings, compared to the complexity inherent to the integration and operations associated with an extra management layer.  

But why play a guessing game, when you can actually calculate it?  Here comes the secret equation which will solve that question once and for all!

Equation

I hope you're not actually trying to prove this equation mathematically.  Instead, note that it's logically divided into 3 segments reflecting the entire rollout of the service delivery:  

Day -1: before service can be actually delivered, hence all the infrastructure preparations  

Day 0: delivery of the service into the network domain  

Day+1: making sure the service is running with the right SLA 

The "variable" constants 

The equation involves a basic set of constants, all related to the service delivery domain in NFV.  First, infrastructure operations (Infra), which represents all the Data-Center’s NFVI/VIM (virtualized infrastructure manager) layers activities associated with getting it ready for service delivery. Second, VNFs, which are the basic functional building blocks of the service. Third, connecting VNFs (Net), which represent the different networking topologies connectivity and service chains, and fourth, configuration (Conf), which addresses the need to define service/user-specific attributes.  

The operators 

The operators involved are multiplication, addition and summation (= SIGMA). But it’s important to use them in the right order!

The addition represents a combination of multiple variants, in our case VNFs, networking topology and configuration, which represents a higher logical entity (i.e. Network Service). The addition of those elements often involves complex logic, such as inter-dependencies and intra-relationships.  The summation is the total addition of all the repetitions (nm) of a given operation. The multiplication operator reflects the need to apply common tasks across all constants/summations, which has a dramatic effect on the final value.   

The "variable" coefficients 

The coefficients in the equation are (Assurance) and (Policy). They serve as a value-multiplier, or factors, that reflect properties that are applied on all constants and variables: monitoring, inventory, topology, troubleshooting, and the enforcement of specific governance on the ways they behave or are used.   

The real variables  

These variables are the most important part of the equation and have a direct and tangible effect on the outcome. (n) represents the number of infrastructure target objects in a given domain. This number can quickly become extremely high in the case of 5G radio. (m) represents the number of times, network services are deployed or updated.   

A practical example  

A given NFV domain consists of 14 datacenters (VIMs) scattered nationwide. The service to be deployed is virtual CPE (customer provided equipment), which is expected to be delivered to over 10,000 enterprise customers.  

So in this case,  n=14 and m=10,000  but that's not all:   

The vCPE itself may be comprised of multiple VNFs, SD-WAN and DC-SDN networking and configuration of end-points. The infrastructure may require multiple prep-operations, such as distribution of assets, creation of tenants, quotas and other artifacts.  There may be a need to police the entire deployment process with anti-affinity and optimization and placement rules. 

These constants and coefficients alone represent BIG numbers. 

So, if you crunch it all together, you undoubtedly get a value outcome that's above "VeryHigh"  

Q.E.D (Quite Easily Demonstrated)!  

Solving the equation  

I'm sure that by now you totally get it.   You get that math was not my most successful class...but the you don’t need to be a mathematician to realize how many times you need to perform an operation and how complex is it. If you need to perform an extremely complex operation once, then it doesn't make sense to invest time in modelling (automating) it, as there's no gain from economies of scale. On the other hand, even if the simplest operation is repeated in large quantities, then it definitely makes sense to automate.  

There's no argument that once n/m are greater than a handful, there are no alternatives to a proper automation system. And clearly, that Excel and batch scripting are not the right tools for ensuring predictable and repeatable service delivery, even at that scale.   

It's not rocket science... just good practical thinking! 

Share your thoughts on this topic by joining the Twitter discussion with @nokia and @nokianetworks using #telcocloud, #cloud, #NFV, #cloudband, #automation  

About Asaf Peled

Asaf is a seasoned engineering professional focusing on networking and cloud architecture, with over 15 years of experience in the telecommunications market. Asaf leads the product management team for Nokia’s CloudBand Network Director, which addresses the NFVO piece in the MANO puzzle.

Tweet me at @peled_asaf