What makes network management so hard? We have lots of metrics, and they’re all less than ideal. So let’s use the single ideal one instead!
The job of network management is about to get even harder. This means you urgently need to change the way you think about, measure and manage networks.
Today’s network delivers us the Web, SaaS applications and Unified Communications. Tomorrow’s will offer us virtual reality, automated homes, smart contacts, and revolutions in energy, health, transport and more.
The ongoing explosion of capacity from fibre and 4G/5G access technologies is increasing the expectation of end users. They demand that everything happens instantaneously, as all networks are now sold as being “superfast”.
Making any application perform is easy if you have an unlimited budget. However, all network managers are under pressure to increase value for money, whilst both delivering on today’s service performance promises and enabling innovative new applications.
This in turn means selecting from a huge menu of application offers, network access technologies, measurements systems, analytics models, configuration choices, and possible workarounds for quality problems.
As a network manager, how can you cut through the bewildering complexity to make the right decisions, for both today’s and tomorrow’s needs?
There is some good news: the more complex a system appears to be, the simpler it actually is. Look at the diagram below. Which is more complex to control: System A or System B?
System A has one degree of freedom and so is a simple as a system can be in terms of control. (It has feedback, so may have complex behaviour.) System B has two degrees of freedom, which is more complex to control, but still manageable. When you get to three degrees of freedom, then you need a magician, and not an engineer!
The more interconnected a system is, the fewer degrees of freedom it has. Consequently, fewer points must be touched (i.e. managed) to impact the whole system.
Holding this in mind, how do we current measure networks? We use a multitude of disconnected (but implicitly coupled) measures: average throughput, delay, jitter, and loss. The result is that we cannot map the amount of capacity – quantity – directly to the user experience – quality – so we over-provision as a workaround.
There is an alternative way of thinking about networks that simplifies this core management task. A single metric exists that connects the end user experience to the complete digital supply chain. This metric captures the quantity of qualityof the packet flow into a single metric “unit” across any network path.
Selecting the right metric greatly simplifies our world, work and job. This single metric is called Quality Attenuation. It is quantified using new mathematics called ∆Q, which unifies packet loss and delay into as a single probabilistic measure.
When this one measure replaces a host of different ones, it massively simplifies the system you are observing and looking to manage:
- Firstly, it splits the network connection latency from the latency caused by IT/cloud systems. This is a huge stride in itself, taking away a common blame game for enterprise IT and ISPs.
- Secondly, it lets you see if the network has sufficient statistical stability. This is called “stationarity” and is an absolute prerequisite to successful management. Get used to hearing this word a lot more in future.
- Thirdly, it allows us to build a robust causal model of any network problems. Is it the WiFi or the fixed network? Is the application being served from close enough to the user? Is poor performance due to a lack of provisioned capacity or poor network quality?
To learn more about how Quality Attenuation and ∆Q metrics can help you to simplify network management, and also to hear industry leading guest speakers, come to our workshop in London on 8th December. Early bird discount ends on Wednesday!
Can you afford the price of coming?
Wait until you find out the cost of ignorance…
For the latest fresh thinking on telecommunications, please sign up for the free Geddes newsletter.
Read It Now
