All Smoke and Mirrors by Michael Kowal – Part 2

…Without The Smoke

 

Welcome back! My last blog “All Smoke and Mirrors Part 1” I left you with several solutions to choose from to propose to ACME's CTO. Read on to learn about the solution I would choose.

 

You decide to suggest that ACME should use solution c, Design #1 and introduce DWDM to maintain an optimal level of redundancy and reduce CAPEX by requiring less physical fiber between data centers. Because there are only two sites, you decide to position a simple, passive DWDM solution. The CTO and Junior (surprise!) are unfamiliar with the solution, so you describe the solution using the following points and simple diagram:

 

  • Instead of using standard “gray” optics in the switches, the customer will use DWDM optics that are tuned to a specific wavelength.
  • The DWDM optics will connect into a passive Add/Drop Multiplexer (also called “filter”), which basically looks like a patch panel. The filter is called ‘passive’ because there is no electricity required for it to function.
  • ACME’s preferred vendor sells the filters in 4-port and 8-port configurations. There are multiple models of each 4 & 8-port filter which each support a range of 40 & 32 wavelengths respectively. For example, 8-port filter ‘model A’ supports wavelengths 1-8, 8-port filter ‘model B’ supports wavelengths 9-16, and so on up to 32.
  • Each port on the filter is statically tuned to a different, specific wavelength. The wavelength of the optic on the switch must match the wavelength of the port you are plugging into on the filter.
  • The filters work in a point-to-point fashion; however multiple filters can be daisy-chained together. If a wavelength is received at a filter and none of its ports match the wavelength, then the wavelength is passed down the daisy-chain (i.e., “expressed”).
  • Each port on the filter is bit-rate agnostic. In other words, the filter uses statically tuned mirrors to pass light - it does not care what transmission speed the router is using (i.e., 1GB or 100GB). In this regard, the passive filter is essentially smoke and mirrors…without the smoke.

 

Blog29-figure1.png

 

As you explain the solution, it becomes clear to the CTO (and Junior) that DWDM reduces the OPEX of leasing additional fibers originally required in Design #1. With DWDM, ACME would only require a single fiber-pair (or, two that traverse separate paths for redundancy). Home-run, right? Of course not! As usual, the CTO will be suspicious of anything as simple as you’ve described and will be looking for any drawbacks. The CTO reveals that there are plans to add two additional data centers, where DC Site 3 and DC Site 4 will each connect directly to DC Site 2, and asks how the DWDM solution will scale to support the additional sites. Regarding the connectivity model, the CTO asks to design DC sites 3 & 4 to connect to each other the same way DC sites 1 & 2 connect to each other and to assume two 400 GB connections: one between one of the spine nodes of DC Site 1 & DC Site 3 and the other between one of the spine nodes of DC Site 2 & DC Site 4.

 

Junior thinks about it for a minute and volunteers to design the DWDM components that are required for the additional data center sites. You almost fall out of your chair with joy; Junior is gaining confidence and becoming self-reliant! All due, of course, to your expert mentorship. The figure below shows the solution Junior came up with.

Blog29-figure2.png

 

Keeping in mind that the filters operate in a point-to-point fashion and will drop wavelengths if the filter has a matching port, Junior finds that he has to map-out all of the wavelengths carefully (sometimes having to cross-connect filters). In the diagram below, Junior whiteboards each site with the required amount of filters (represented by the trapezoid icons, each containing the range of wavelengths supported) to support all required connections while only using a single fiber pair between each site.

Blog29-figure3.png

The CTO is very pleased to see Junior’s final design on the whiteboard because, in terms of operational simplicity, it is a complete nightmare. The CTO points out that, while the solution does leave some room for additional connectivity, tracing circuits between equipment while troubleshooting will be very difficult; the last thing that his team needs to be doing during an outage is trying to understand how device connectivity is mapped through the passive DWDM system.

 

At this point, you can sense the CTO’s frustration. From his perspective, he feels as if he is being taken down the wrong path and his time is being wasted. Before he begins to reset, you have a suggestion that would significantly reduce the complexity of the passive DWDM design while maintaining optimal redundancy: introduce a Reconfigurable Optical Add-Drop Multiplexer (ROADM) to the solution.

 

You explain to the CTO that with a passive filter, different wavelengths are statically mapped across ports on a filter and due to the filter only operating in a point-to-point fashion, Junior had to keep track of where wavelengths were being added and dropped. For example, assume we use channel 1 on a filter and daisy-chain multiple filters together. Once we encounter another filter in the chain that used channel 1, the filter would drop it. Now, with a ROADM, we can receive channels at a site and, in software, dynamically select which channels we want to add/drop or pass on (i.e., express). We no longer need to keep track of which channels are added/dropped at each site, thus simplifying the physical design. Also, ROADMs are multi-degree capable (they can send/receive wavelengths in more than two directions) where the filter was only capable of only two directions. Finally, you mention that the active DWDM system is capable of transporting up to 96 wavelengths which provides greater scalability over the passive DWDM option.

 

The CTO sounds interested and requests a whiteboard drawing of the concept.Blog29-figure4.png

After some consideration, it seems like the CTO approves of the design and we’re good to go. However, and as usual, the CTO simply mentions that he’ll bring this to his team and that he’ll get back to Junior after they’ve reviewed the design internally. As your eyes roll all the way in the back of your head - you thank him for his time, hang up the phone, and realize your day is almost over with nothing accomplished on your to-do list. Maybe it’s for the best that you never leave the home office again…

 

Was the solution c the one you chose on Part 1? It’s alright if not, as different network designers have different approaches to meet the same business and technical requirements, no right or wrong in this case. Hope you enjoyed this 2-part article and mainly that you learned a little bit about DWDM.

 

About the Author

pic Michael Kowal.jpg

For the past 12 years, Michael Kowal has been involved with carrier routing and optical designs and architectures at Cisco. He currently works with national research institutions, regional service providers, large-scale government & higher education customers to help educate, design, and architect Evolved Programmable Networks. Michael's technology focus within the Public Sector includes: BGP, LISP, Segment Routing, IPv6 and DWDM.

Michael currently holds a CCDE and a CCIE in the Routing & Switching, Service Provider, and Voice tracks. Michael also holds a Masters in Electrical Engineering from Stevens Institute of Technology.

 

Here are a few additional ways for us to engage and keep the conversation going: