Jay, Aria Networks CTO, explains why Service Awareness and Sensitivity Analysis need to be at the top of agenda when planning capacity and network change.
With a history of research, development and support I’ve had my fair share of experience with the technical aspects of telco networks: From the elegant structure of the TDM hierarchy through managing optical impairments and IP packet queuing. Such immersion in technology can remove you from what telco networks are really about. It is not about a cool way of transporting Ethernet transparently across an optical network. It is about generating revenue from the provision of services.
Designing a network to deliver services without a clear understanding of those services or even a traffic matrix is to set yourself up for fall. Obviously there are times when the profile of services is unknown but this is not a reason to focus on technology but to be clear and determine what good really looks like.
I am not saying that standards and protocols are not important. What I am saying, though, is when using planning tools to design a network architecture those tools must take into account the services being transported or to be more precise the services should be used to drive the architecture.
So, let’s just assume for now that it is possible to design a network using the services as the input and use revenue, or to be more precise margin, as the goal. What happens if you get the traffic matrix, forecast or service mix wrong? Well the answer comes back to a previous blog: ‘what does good look like?’. In this instance what good looks like is an estimate of the architecture with bounds which are a result of the service uncertainty.
[DA: Ok you’re losing me here. How on earth can you estimate an architecture and even more bizarrely suggest bounds?]
I appreciate this may be a little odd to think of in this way but in reality if we think about the network as a calculator we have at our disposal a number of options; where protection is done, where switching is performed, the type of aggregation, vendor specific topologies or even creation of currently non-commercially available equipment. If ‘good’ means the delivery of services in a cost efficient way then a good planning tool should combine, manipulate, trade off and dimension these options such that we get the required architecture. Furthermore it’s an estimate because there are uncertainties in the service parameters.
An important step in the modelling process is to perform a sensitivity analysis, like a Monte Carlo, to determine the bounds of a solution. To ground this in reality the bounds may be something as simple as the utilisation bounds to something more dramatic like a cost efficient solution that could embrace different architectures like MPLS TP or layer 3 switching and an OTN solution. I have experienced modelling a tier one network where the static service profile suggests a comfortable link utilisation but when a realistic sensitivity analysis is done a much more alarming picture materialises.
The objective of this blog is to suggest when planning a network, be it strategic or tactical, it is important to put Service Awareness and Sensitivity Analysis at the top of modelling agenda. After all, network planning is all about the services.
The ultimate vision here is using a planning tool which takes as its input the services required, desired margin and measurement of uncertainty. The planning tool would then be used to ‘Goal Seek’; manipulating the network so that the margin is delivered within the bounds of service uncertainty. How cool would that be?