Q&A with Shannon McFarland, Co-Author of IPv6 for Enterprise Networks
Like this article? We recommend
Like this article? We recommend
Like this article? We recommend
Introduce yourself and give us a bit of background on you professionally.
I am Shannon McFarland, and I work on the Consulting Engineering team in the Office of the CTO at Cisco. I have been at Cisco 11+ years focused on IPv6 as well as data center design. Most of my time is spent working to help customers understand how to design their enterprise networks to support IPv6 and their data center to support specific applications, such as virtual desktop infrastructure (VDI) or Microsoft server applications. I have for many years focused on the validation of designs and authoring Cisco Validated Designs (CVDs). I still do that stuff, but my focus now includes interfacing with the customer and Cisco’s internal product teams to ensure we are building stuff they need sooner rather than later; requirements gathering sort of stuff.
Prior to Cisco, I worked at a systems integrator for several years and within IT in the healthcare industry before that. I have always had a mix of network and application deployment in my blood.
Why did you write this book and why now?
Well, I co-authored this book with some other Cisco folks, Muninder Sambi, Nikhil Sharma, and Sanjay Hooda, who each have a unique history regarding IPv6 and networking in general, which includes advanced services, product management, and development.
We wrote this book because we knew enterprise customers all over the world were already deploying IPv6 and needed more guidance than we could give them 1-to-1. I wrote two Cisco Validated Designs for Campus and Branch IPv6 deployment back in 2009, and those have been used all over the place. That work was the basis for this book. As a guy who lives and breathes deployment in the enterprise space, I knew now was the time to get something out to help those in early or mainstream deployment.
As for the “why now” question, I have been seeing a trend for the last 2-3 years of enterprise customers large and small who were planning their deployments for 2011-2012. This really had zero to do with the IPv4 address exhaustion pandemonium you have been reading about in blogs and tweets. I have supported enterprise IPv6 deployments since the early 2000s, and it had nothing to do with panicking over IANA or an RIR running out of addressing. It was all internal use case stuff. Let me go out on a tangent for a second. People ask me, “Why do people assume enterprises aren’t deploying IPv6?” The reality is that they have been deploying and for sometime but they have zero justification for telling anyone they have. The enterprise is radically different than the SP space where the SP needs people to know they can support IPv6. They do this to keep their existing customers and draw new ones away from the lagging SPs. Enterprises, with the exception of vendors who sell gear that is IP-aware, have no need to market their fancy new IPv6-enabled network. In fact, I have had customers who were told by their shareholder relations teams to not tell anyone they were doing IPv6, as it may be perceived as overly risky. Now, people are coming out and talking about being IPv6-enabled, to let their customer base know they are prepared for whatever happens with IPv4 address exhaustion, and that they are future-proofing their network. It is bizarre how things change.
Anyhow, people kept telling me, “You should get this stuff in a book, as I know we aren’t the only customer asking you this stuff.” So, here we are.
What concerns you about this book and enterprise IPv6 deployment in general?
What concerns me about any book, not just this one, is being thorough enough without making the details so specific that they become irrelevant by the time a reader gets the book in their hands. This book is no different.
One regret I have with this book, and any others I have read from any author on the topic of IPv6, is deep dark design recommendations and best practices around exactly what to do on the internet edge (connection between the enterprise and their providers). We touched a bit on it but quite frankly, we did not offer as much as people need[md]not even close. This was not something that happened by accident. What we did not want to do is propose a few designs that we would consider to be mainstream designs and then end up finding out from key customers that we missed the boat entirely. The stuff in this book is not experimental or built only in a lab, but is based squarely on truckloads of real deployments. So, leaping out on the edge (no pun intended) and talking about detailed internet edge designs that really are not well baked in the ISP space is just not what we wanted to do.
There are numerous enterprise customers running production-quality IPv6 today and have been. They are also on the Internet with their IPv6 connections, but I have never seen two do it the same way. There are many theoretical designs and even semi-safe practical ones, but to confidently say “these one or two designs are the way everyone should do it” is ridiculous at this point. There are many ways and they depend on the specific customer environment, their technical capabilities (i.e. network products, applications), business justification, timeframe, and certainly the capabilities of their ISP(s).
Ideally the topic of multihoming should be similar, if not exactly the same between IPv4 and IPv6, but the reality is that it isn’t that way in every occasion, at least not today. We have customers that have to tunnel between an edge router across their normal IPv4 POP connection to an IPv6-enabled POP in another city because the provider has not dual-stacked all of their POPs. Stuff like that drives you crazy and causes internet edge routing designs to not be exactly the same or even similar all of the time.
What do you think is unique about this book?
It is solely about the enterprise. It has very little theory and IPv6 101-type content, which we will get dinged for by some, but the reality is that there is so much 101-level content out there that you can get for free. Burning pages in a book for it is just not worth the reader’s time or money, as we would have had to sacrifice page count from deployment stuff to make room for primer stuff.
I do deployment; the book is mostly about deployment. Does it fit every scenario? Nope. No book could cover all of the possibilities, but it does offer some guidance that is based on years of real deployments, lessons learned and stuff that will kill you in a real implementation.
Folks, there is still a boatload of stuff we just don’t know yet about IPv6 deployment. Sure, we have lots of deployments out there and we have learned tons from them, but we still have a ways to go. How well will it go when you try to reach an enterprise-hosted web site from an IPv6 only mobile device? Did they use NAT64? A proxy? A load balancer with translation functionality? Did DNS do what it is supposed to do? Why can’t Apple get DHCPv6 client support in their operating system so I can get rid of Stateless Address Autotconfiguration (SLAAC) and just use DHCP? All of these are things that will change over time and will be customer-dependent. We are just scratching the surface of what an enterprise has already and/or will face during their IPv6 deployment.